# everyais API and public data reference

<a id="answer-first"></a>
## Answer first

everyais is an AI API gateway run by BROZ Corp. in Korea. One API key calls Claude, GPT, Gemini, DeepSeek and the other models in its public catalog through OpenAI- and Anthropic-compatible endpoints. Every model has a public price, and each key can have its own spend limit.

Updated: 2026-10-28T07:25:33.035Z

<a id="key-facts"></a>
## Key facts

- APIs: OpenAI Chat Completions (`/v1/chat/completions`) and Responses (`/v1/responses`), Anthropic Messages (`/v1/messages`), image generation and editing (`/v1/images/generations`, `/v1/images/edits`), and video generation (`/v1/videos/generations`). Existing OpenAI and Anthropic SDKs work by changing the base URL and API key.
- Billing: usage is deducted from prepaid credits at the public USD reference prices, with no subscription fee. New accounts get free credit to try the API.
- Coding tools: Claude Code, Cursor and opencode connect through the same API. The docs include settings for each.
- Fallback: when a model runs on several providers, a failed request moves to another one. Requests can also name up to 5 backup models.
- Cost controls: each API key can restrict allowed models and set request limits and daily or monthly spend limits.
- Training: routes where the provider trains on prompts are off unless the account opts in.
- Data handling: requests are sent to the inference route of the selected model; retention and provider terms are listed in the privacy policy.
- Policies: https://platform.everyais.com/en/privacy, https://platform.everyais.com/en/terms, https://platform.everyais.com/en/refund
- Contact: portal@everyais.com

<a id="definitions"></a>
## Definitions

- Reference price: the public catalog price, including model discounts, used for comparison; account and organization rates can change the final billed price.

<a id="model-catalog"></a>
## Model catalog

Browse the callable models and their reference prices on the public pages below.
- English: https://platform.everyais.com/en/models
- 한국어: https://platform.everyais.com/models
- [GPT-6.1 Terra](https://platform.everyais.com/en/models/everyais/gpt-6-1-terra): `everyais/gpt-6-1-terra`; Standard (≤272K): Input $1 · Output $5 /1M · Long context (>272K, full request): Input $2 · Output $7.5 /1M
- [Gemini 4 argon](https://platform.everyais.com/en/models/everyais/gemini-4-argon): `everyais/gemini-4-argon`; text, image, video → text; 1.0M context; Input $1.4 · Output $7 /1M
- [Claude Haiku 5.5](https://platform.everyais.com/en/models/everyais/claude-haiku-5-5): `everyais/claude-haiku-5-5`; 1.0M context; Input $0.1 · Output $0.5 /1M
- [Gemini nano banana 2.1](https://platform.everyais.com/en/models/everyais/gemini-nano-banana-2-1): `everyais/gemini-nano-banana-2-1`; text, image, video → text, image; Input $1.5 · Output $30 /1M
- [GPT-6.1 Sol](https://platform.everyais.com/en/models/everyais/gpt-6-1-sol): `everyais/gpt-6-1-sol`; text → text; 1.1M context; Standard (≤272K): Input $2 · Output $10 /1M · Long context (>272K, full request): Input $4 · Output $15 /1M
- [Grok 4.7](https://platform.everyais.com/en/models/everyais/grok-4-7): `everyais/grok-4-7`; Input $2 · Output $6 /1M
- [Claude Sonnet 5.5](https://platform.everyais.com/en/models/everyais/claude-sonnet-5-5): `everyais/claude-sonnet-5-5`; text, image → text; 1.0M context; Input $2 · Output $10 /1M
- [GPT-6 Luna](https://platform.everyais.com/en/models/everyais/gpt-6-luna): `everyais/gpt-6-luna`; text, image → text; 1.1M context; Standard (≤272K): Input $0.1 · Output $0.5 /1M · Long context (>272K, full request): Input $0.2 · Output $0.75 /1M
- [GPT-6 Sol](https://platform.everyais.com/en/models/everyais/gpt-6-sol): `everyais/gpt-6-sol`; text, image → text; 1.1M context; Standard (≤272K): Input $2 · Output $10 /1M · Long context (>272K, full request): Input $4 · Output $15 /1M
- [Claude Opus 5.5](https://platform.everyais.com/en/models/everyais/claude-opus-5-5): `everyais/claude-opus-5-5`; text, image → text; 1.0M context; Input $4 · Output $20 /1M
- [GPT-6 Astra](https://platform.everyais.com/en/models/everyais/gpt-6-astra): `everyais/gpt-6-astra`; text, image → text; 1.1M context; Standard (≤272K): Input $10 · Output $50 /1M · Long context (>272K, full request): Input $20 · Output $75 /1M
- [GPT Image 2.5 Sunburst](https://platform.everyais.com/en/models/everyais/gpt-image-2-5-sunburst): `everyais/gpt-image-2-5-sunburst`; text, image → image; Input $5 · Output $30 /1M
- [GPT Image 2.5 Flare](https://platform.everyais.com/en/models/everyais/gpt-image-2-5-flare): `everyais/gpt-image-2-5-flare`; text, image → image; Input $5 · Output $30 /1M
- [Gemini 3.8 Flash](https://platform.everyais.com/en/models/everyais/gemini-3-8-flash): `everyais/gemini-3-8-flash`; text, image → text; 1.0M context; Input $0.525 · Output $2.625 /1M
- [Claude Fable 5.1](https://platform.everyais.com/en/models/everyais/claude-fable-5-1): `everyais/claude-fable-5-1`; text, image → text; 1.0M context; Input $10 · Output $50 /1M
- [Kimi K3](https://platform.everyais.com/en/models/everyais/kimi-k3): `everyais/kimi-k3`; text → text; 1.0M context; Input $3 · Output $15 /1M
- [Gemini Omni 1.1 Flash](https://platform.everyais.com/en/models/everyais/gemini-omni-1-1-flash): `everyais/gemini-omni-1-1-flash`; text → video; Input $1.5 · Output $17.5 /1M
- [Nemotron 3.5 Lightning](https://platform.everyais.com/en/models/everyais/nemotron-3-5-lightning): `everyais/nemotron-3-5-lightning`; text → text; 1.0M context; Input $0.06 · Output $0.24 /1M
- [Gemini Omni Flash Preview](https://platform.everyais.com/en/models/everyais/gemini-omni-flash-preview): `everyais/gemini-omni-flash-preview`; text → video; Input $1.5 · Output $17.5 /1M
- [Gemini 3.1 Flash Lite Image](https://platform.everyais.com/en/models/everyais/gemini-3-1-flash-lite-image): `everyais/gemini-3-1-flash-lite-image`; text → image; Input $0.25 · Output $30 /1M
- [Claude Fable 5](https://platform.everyais.com/en/models/everyais/claude-fable-5): `everyais/claude-fable-5`; text, image → text; 1.0M context; Input $10 · Output $50 /1M
- [MiniMax M3](https://platform.everyais.com/en/models/everyais/minimax-m3): `everyais/minimax-m3`; text → text; 1.0M context; Input $0.3 · Output $1.2 /1M
- [Gemini 3.1 Flash Image](https://platform.everyais.com/en/models/everyais/gemini-3-1-flash-image): `everyais/gemini-3-1-flash-image`; text → image; Input $0.5 · Output $60 /1M
- [DeepSeek V4 Pro 0813](https://platform.everyais.com/en/models/everyais/deepseek-v4-pro-0813): `everyais/deepseek-v4-pro-0813`; text → text; 979K context; Input $1.32 · Output $3.96 /1M
- [DeepSeek V4 Flash 0731](https://platform.everyais.com/en/models/everyais/deepseek-v4-flash-0731): `everyais/deepseek-v4-flash-0731`; text → text; 1.0M context; Input $0.14 · Output $0.28 /1M
- [Gemini 3.1 Pro Preview](https://platform.everyais.com/en/models/everyais/gemini-3-1-pro-preview): `everyais/gemini-3-1-pro-preview`; text, image → text; 1.0M context; Standard (≤200K): Input $2 · Output $12 /1M · Long context (>200K, full request): Input $4 · Output $18 /1M
- [GLM-5.3](https://platform.everyais.com/en/models/everyais/glm-5-3): `everyais/glm-5-3`; text → text; 1.0M context; Input $1.4 · Output $4.4 /1M
- [GLM-5.3 Flash](https://platform.everyais.com/en/models/everyais/glm-5-3-flash): `everyais/glm-5-3-flash`; text, image → text; 1.0M context; Input $0.15 · Output $0.5 /1M
- [Gemini 3 Pro Image](https://platform.everyais.com/en/models/everyais/gemini-3-pro-image): `everyais/gemini-3-pro-image`; text → image; Input $2 · Output $120 /1M
- [Gemini 2.5 Flash Image](https://platform.everyais.com/en/models/everyais/gemini-2-5-flash-image): `everyais/gemini-2-5-flash-image`; text → image; Input $0.3 · Output $30 /1M
- [Qwen 3.8 27B](https://platform.everyais.com/en/models/everyais/qwen-3.8-27b): `everyais/qwen-3.8-27b`; text, image, video → text; 262K context; Input $0.45 · Output $3 /1M
- [MiniMax M3.1 Flash](https://platform.everyais.com/en/models/everyais/minimax-m3-1-flash): `everyais/minimax-m3-1-flash`; 1.0M context; Input $0.15 · Output $0.6 /1M
- [DeepSeek V4.1 Flash 1017](https://platform.everyais.com/en/models/everyais/deepseek-v4-1-flash-1017): `everyais/deepseek-v4-1-flash-1017`; text, image → text; 1.0M context; Input $0.075 · Output $0.3 /1M
- [Muse Spark Open](https://platform.everyais.com/en/models/everyais/muse-spark-open): `everyais/muse-spark-open`; Input $1.25 · Output $4.25 /1M
- [Nemotron 3.5 Super VL](https://platform.everyais.com/en/models/everyais/nemotron-3-5-super-vl): `everyais/nemotron-3-5-super-vl`; Input $0.3 · Output $0.8999999999999999 /1M
- [FLUX 3 Image Klein](https://platform.everyais.com/en/models/everyais/flux-3-image-klein): `everyais/flux-3-image-klein`; $0.009 /image (768 $0.008 · 1K $0.009 · 1.5K $0.011 · 2K $0.014)
- [MiniMax H3 Turbo](https://platform.everyais.com/en/models/everyais/minimax-h3-turbo): `everyais/minimax-h3-turbo`; text, image → video; $0.03 /sec

<a id="api-docs"></a>
## API docs

- Base URL: https://api.everyais.com/v1
- Authentication: `Authorization: Bearer everyais_...`
- OpenAPI: https://api.everyais.com/openapi.json
- English: https://platform.everyais.com/en/docs
- 한국어: https://platform.everyais.com/docs

# Getting started

## Quickstart

_Send your first request with curl, Python, or Node.js._

Source: https://platform.everyais.com/en/docs/quickstart
Updated: 2026-10-22T12:28:27.013Z

## 1. Get an API key

[Create a key](https://platform.everyais.com/dashboard/keys) and [add credits](https://platform.everyais.com/dashboard/pricing). The key is shown only once, so save it as an environment variable.

```bash
export EVERYAIS_API_KEY="everyais_your_api_key"
```

## 2. Choose a model

```bash
export EVERYAIS_MODEL="everyais/gemini-3-8-flash"
```

Copy other IDs from the [model directory](https://platform.everyais.com/en/models). `GET /v1/models` lists the models your key can call; see [model selection](https://platform.everyais.com/en/docs/model-selection).
A key with scopes needs `models:read` to list models and `chat` to generate text.

## 3. Send your first request

### curl

```bash
curl https://api.everyais.com/v1/chat/completions \
  -H "Authorization: Bearer $EVERYAIS_API_KEY" \
  -H "Content-Type: application/json" \
  -d @- <<EOF
{
  "model": "$EVERYAIS_MODEL",
  "messages": [{"role": "user", "content": "Hello!"}]
}
EOF
```

### Python

```bash
pip install openai
```

```python
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.everyais.com/v1",
    api_key=os.environ["EVERYAIS_API_KEY"],
    timeout=240.0,
)
response = client.chat.completions.create(
    model=os.environ["EVERYAIS_MODEL"],
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)
```

### Node.js

```bash
npm install openai
```

```javascript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.everyais.com/v1",
  apiKey: process.env.EVERYAIS_API_KEY,
  timeout: 240_000,
});
const response = await client.chat.completions.create({
  model: process.env.EVERYAIS_MODEL,
  messages: [{ role: "user", content: "Hello!" }],
});
console.log(response.choices[0].message.content);
```

## Next steps

| Goal | Documentation |
|------|---------------|
| Show tokens as they arrive | [Streaming](https://platform.everyais.com/en/docs/streaming) |
| Connect a coding tool | [Claude Code](https://platform.everyais.com/en/docs/integration-claude-code) · [Cursor](https://platform.everyais.com/en/docs/integration-cursor) · [opencode](https://platform.everyais.com/en/docs/integration-opencode) |
| Use the Anthropic SDK | [Messages](https://platform.everyais.com/en/docs/messages) |
| Run long text generation | [Async jobs](https://platform.everyais.com/en/docs/async-jobs) |
| Generate or edit images | [Image generation](https://platform.everyais.com/en/docs/images-generations) · [Image editing](https://platform.everyais.com/en/docs/images-edits) |
| Generate video | [Video jobs](https://platform.everyais.com/en/docs/videos-generations) |
| Handle a failed request | [Errors and retries](https://platform.everyais.com/en/docs/errors-and-rate-limits) |

SDK references: [Python SDK](https://github.com/openai/openai-python#usage) · [Node.js SDK](https://github.com/openai/openai-node#usage).

---

## Base URL

_The base URL for each client, timeouts, and request size limits._

Source: https://platform.everyais.com/en/docs/base-url
Updated: 2026-10-22T12:28:16.529Z

## Base URL by client

| Client | Base URL |
|--------|----------|
| OpenAI SDK, Cursor, opencode and other OpenAI-compatible tools | `https://api.everyais.com/v1` |
| Anthropic SDK, Claude Code | `https://api.everyais.com` |

OpenAI-compatible clients append paths such as `/chat/completions`, so include `/v1`. The Anthropic SDK adds `/v1/messages` itself, so leave it out.
A doubled prefix such as `/v1/v1/messages` returns 404.

## Timeouts

Response time depends on the model and input size. Set your own client timeout, and check your usage before you retry a request that timed out.
For long tasks, use [streaming](https://platform.everyais.com/en/docs/streaming) or [async jobs](https://platform.everyais.com/en/docs/async-jobs).

## Request body size

| Request | Limit |
|---------|-------|
| Chat, Messages, Responses | 5 MiB |
| Image editing, multipart (whole body) | 4 MiB |
| Image editing, JSON | 5 MiB |
| Video generation | 5 MiB |
| Everything else, including image generation | 1 MiB |

Larger bodies return 413 `request_too_large`. Multipart limits include boundaries and form fields.

## Public resources

The [OpenAPI spec](https://api.everyais.com/openapi.json), [documentation index](https://platform.everyais.com/llms.txt) and [full documentation](https://platform.everyais.com/llms-full.txt) need no authentication.
Browse the [model directory](https://platform.everyais.com/en/models), then call `GET /v1/models` to see which models your key can use.

---

## Authentication

_API keys, auth headers, and per-key limits and scopes._

Source: https://platform.everyais.com/en/docs/authentication
Updated: 2026-10-22T12:28:21.772Z

Send your API key in the `Authorization` header on inference and `/v1/models` requests.

```
Authorization: Bearer everyais_your_api_key
```

Create keys in the [dashboard](https://platform.everyais.com/dashboard/keys). We store only a hash, so the full key is shown once. Save it right away.
The Anthropic SDK's `x-api-key: everyais_...` header also works. If you send both, a valid Bearer header wins.
Keep keys on your server. Never put them in browser code or public repositories.

## Per-key settings

| Setting | Behavior |
|---------|----------|
| Rate limit | 100 requests per minute by default. Change it when you create or edit the key |
| Spend limit (USD) | Monthly, daily and per-minute. Over the limit returns 402 `spend_limit_exceeded` |
| Allowed models | Other models return 403, and `GET /v1/models` lists only these |
| Scopes | `chat`, `images`, `video`, `models:read`. Calls outside them return 403. No scopes allows everything |
| Expiry date | Requests after this date return 401 |

## What keys can't do

Account settings, payments and key management need a signed-in session. An API key gets 403 `api_key_session_forbidden`.
Supplier accounts can't use the APIs in these docs (403 `account_kind_forbidden`).

Organization keys use the organization's credits and settings, including community supply and training use. See [model selection and access](https://platform.everyais.com/en/docs/model-selection).

---

## Model selection and access

_Model IDs, capabilities, parameter policy, and account or organization access settings._

Source: https://platform.everyais.com/en/docs/model-selection
Updated: 2026-10-22T12:30:21.308Z

## 1. Find models available to your key

`GET /v1/models` returns the models your key can call, after its allowed-model list and your account or organization settings.
The public [model directory](https://platform.everyais.com/en/models) also shows models that need consent. Use the exact `id` from the API.

## 2. Check capabilities

Check `category`, `supported_endpoints`, `available`, `capabilities` and `limits` before you send a request. A missing `capabilities` or `limits` field means support is not confirmed.
Model limits can be lower than the gateway's general limits. Messages and Responses are translated to Chat, so they don't expose every provider feature.
See [model fields](https://platform.everyais.com/en/docs/models#doc-model-fields) for every field.

### Parameter policy

Use the selected model’s `parameter_profile` for `required` inputs, `defaults` for omitted values, supported `parameters` and allowed `combinations`.
Set the policy in the request body, for example `"everyais": {"parameter_policy": "strict"}`.

| `everyais.parameter_policy` | Behavior |
|---|---|
| `auto` (default) | Drops unsupported sampling options (`temperature`, `top_p`, penalties, `top_k`), `seed` and `user`, and lowers values to a supported one. Other unsupported options, such as `stop`, `logprobs` or tools, return **400** |
| `strict` | Any unsupported value or combination returns **400** `unsupported_parameter` |

Each adjustment is reported in `x_everyais.parameter_warnings` with `param`, `action`, `reason` and `applied`.
Tools, masks and response formats are never adjusted. Options that only some providers support are listed in `parameter_constraints` and `parameter_profile.routes`.

## 3. Check access policies

Community supply (`supply`) and provider training use (`training_use`) are off until you opt in. Personal keys use your account settings; organization keys use the organization's.

A model available only through community supply or only with training use returns 404 `model_not_found` on Chat, Messages and Responses until you opt in. Requests to `mixed` models go only to providers you allow.
If a model is missing, check the key's allowed models, both settings and the model's `available` flag.

---

# Endpoints

## POST /v1/chat/completions

_OpenAI-compatible chat completions with streaming, tool calls, and structured outputs._

Source: https://platform.everyais.com/en/docs/chat-completions
Updated: 2026-10-25T17:43:53.915Z

Creates a model response for a conversation, in the OpenAI Chat Completions format.

## Request

```json
{
  "model": "everyais/gemini-3-8-flash",
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Hello!"}
  ],
  "stream": false
}
```

## Response

```json
{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1709884800,
  "model": "everyais/gemini-3-8-flash",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello! How can I help you today?"
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 25,
    "completion_tokens": 10,
    "total_tokens": 35
  }
}
```

## Parameters

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `model` | string | Yes | Model ID, such as `everyais/gemini-3-8-flash` |
| `messages` | array | Yes | 1–1000 messages. `role`: `system`, `user`, `assistant`, `tool` or `developer` (treated as `system`). `content`: a string, null, or up to 1000 parts. Send images (`image_url`) as base64 data URLs. Only open-weight models (vLLM-based, such as Nebius) accept `https` URLs; Claude, GPT and Gemini return **400** for them |
| `stream` | boolean | No | Stream server-sent events. Default `false` |
| `stream_options` | object | No | `{"include_usage": true}` adds a final usage chunk |
| `temperature` | number | No | 0–2 |
| `top_p` | number | No | 0–1 |
| `stop` | string \| string[] | No | Up to 4 sequences |
| `n` | integer | No | Only `1`. 2 or more returns 400 |
| `presence_penalty` / `frequency_penalty` | number | No | -2–2 |
| `seed` | integer | No | Best-effort reproducibility |
| `tools` | array | No | Up to 512 function tools, plus one `{"type":"web_search"}`. See [web search](https://platform.everyais.com/en/docs/web-search) |
| `tool_choice` | string \| object | No | `auto`, `none`, `required` or `{"type":"function","function":{"name":"..."}}`. `none` also disables web search. Some models (Claude Opus 5.5, Sonnet 5.5, Fable 5.1) return **400** for forced tool use; see `tool_choice` in `parameter_profile` |
| `response_format` | object | No | `{"type":"text"}`, `{"type":"json_object"}` or `{"type":"json_schema","json_schema":{...}}` |
| `reasoning_effort` | string | No | `low`, `medium`, `high`, `xhigh`, `max`. Some models add values such as `none` or `minimal` |
| `parallel_tool_calls` | boolean | No | Allow parallel tool calls. `true` is the default and equals omitting it. `false` works only on models whose `capabilities.parallel_tool_calls` is `true` |
| `logprobs` / `top_logprobs` | boolean / integer | No | Token log probabilities. `top_logprobs`: 0–20 |
| `verbosity` | string | No | `low`, `medium`, `high`. GPT models only; others return **400** |
| `prompt_cache_key` | string | No | OpenAI prompt cache key, up to 1024 characters. Used by GPT models only; other models drop it with a warning, or return **400** with `parameter_policy: "strict"` |
| `safety_identifier` | string | No | Hashed end-user ID for abuse detection, up to 64 characters. GPT models only |
| `logit_bias` | object | No | Token ID to bias (-100–100), up to 300 entries. Open-weight models only |
| `user` | string | No | End-user identifier |
| `everyais` | object | No | Gateway options: `cache`, `cache_ttl`, `parameter_policy`, `provider`. See [prompt caching](https://platform.everyais.com/en/docs/prompt-caching) |
| `provider` | object | No | Provider preferences: `sort` (`price`, `latency`, `throughput`) and `allow_fallbacks`. You can't pick a provider by name |
| `models` | string[] | No | Fallback models, up to 5. Tried in order if the first model is unavailable |
| `extra_body` | object | No | Only `anthropic`, `google`, `openai` and `everyais` keys; others return 400. Gemini takes `google.top_k` and `google.media_resolution` (`low`, `medium`, `high`) |

Model ID suffixes set provider order: `:nitro` for lowest latency, `:floor` for cheapest provider first. Your price stays the same. A reasoning effort suffix such as `:high` sets `reasoning_effort` when the body leaves it out, and it combines with a speed suffix (`:high:nitro`). Use it in tools that can't send reasoning effort, such as Cursor.
Unknown fields such as `store` and `service_tier` return **400**.

## Model options

Pick a model at the top of this page, or read `capabilities`, `limits` and `parameter_constraints` from `GET /v1/models`. Use only listed `allowed_values`.
[Parameter policy](https://platform.everyais.com/en/docs/model-selection#doc-parameter-policy) decides whether unsupported values are adjusted or rejected with **400**.

- Strict JSON schemas need `capabilities.structured_outputs: true`.
- To continue a tool conversation, keep the same model and send back the original `tool_calls` unchanged. Gemini needs `tool_calls[].extra_content.google.thought_signature`; OpenAI models need `extra_content.openai.reasoning_id` and `reasoning_encrypted_content`.

## Streaming

With `stream: true`, the response is `data: {...}` events ending with `data: [DONE]`. See [streaming](https://platform.everyais.com/en/docs/streaming).

---

## POST /v1/messages

_Anthropic Messages-compatible endpoint, plus free token counting._

Source: https://platform.everyais.com/en/docs/messages
Updated: 2026-10-22T12:28:37.498Z

Creates a message in the Anthropic Messages format. The Anthropic SDK and Claude Code work with base URL `https://api.everyais.com`.

## Request

```json
{
  "model": "everyais/claude-sonnet-5-5",
  "messages": [{"role": "user", "content": "Hello!"}]
}
```

## Response

```json
{
  "id": "msg_example",
  "type": "message",
  "role": "assistant",
  "model": "everyais/claude-sonnet-5-5",
  "content": [
    {
      "type": "text",
      "text": "Hello! How can I help you today?"
    }
  ],
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": {
    "input_tokens": 25,
    "output_tokens": 10
  }
}
```

## Parameters

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `model` | string | Yes | Model ID |
| `messages` | array | Yes | `role`: `user` or `assistant`. `content`: a string or blocks (`text`, `image`, `tool_use`, `tool_result`, `thinking`, `document`) |
| `system` | string \| array | No | A string or text blocks |
| `temperature` | number | No | 0–1 |
| `top_p` | number | No | 0–1 |
| `top_k` | integer | No | Positive integer |
| `stop_sequences` | string[] | No | Up to 8 |
| `stream` | boolean | No | Default `false` |
| `tools` | array | No | `{name, description, input_schema}` |
| `tool_choice` | object | No | `{"type":"auto"\|"any"\|"none"}` or `{"type":"tool","name":"..."}`. Models that reject forced tool use in Chat also reject `any` and `tool` |
| `metadata` | object | No | `{"user_id": "..."}`, passed through as Anthropic metadata |
| `thinking` | object | No | `{"type":"enabled","budget_tokens":N}` (N ≥ 1024), `{"type":"adaptive"}` or `{"type":"disabled"}` |
| `output_config` | object | No | `effort` (`low`–`max`, with adaptive or omitted thinking) and `format` (`{"type":"json_schema","schema":{...}}`, structured-output models only) |
| `everyais` | object | No | Gateway options such as caching and `parameter_policy` |

## Model options

Pick a model at the top of this page, or read `capabilities`, `limits` and `parameter_constraints` from `GET /v1/models`.
[Parameter policy](https://platform.everyais.com/en/docs/model-selection#doc-parameter-policy) decides whether unsupported values are adjusted or rejected with **400**.

- `document` blocks and `tool_result` blocks with images return **400**.
- Send images with `source.type: "base64"`. `url` sources follow the [Chat](https://platform.everyais.com/en/docs/chat-completions) image URL rules.
- `cache_control` on a block is passed through unchanged. See [prompt caching](https://platform.everyais.com/en/docs/prompt-caching).
- To continue a tool conversation, send the response `content` blocks back unchanged, including `thinking` blocks. Gemini's tool signature is in the `thinking` block's `signature`.
- `stop_reason: "stop_sequence"` and `stop_sequence` are returned when the provider reports the matched string (Claude and vLLM-based models). Gemini and GPT return `end_turn`.
- Web search is available only in Chat.

---

## POST /v1/messages/count_tokens

Counts input tokens for free. Claude Code calls it automatically. The body is the same as `/v1/messages`, and the count is an estimate.

```json
{
  "input_tokens": 1234
}
```

---

## POST /v1/responses

_Responses format, background jobs, and cancellation._

Source: https://platform.everyais.com/en/docs/responses
Updated: 2026-10-25T17:44:00.209Z

Creates a response in the OpenAI Responses format. Conversations are not stored, so send earlier turns in `input`.

## Request

```json
{
  "model": "everyais/gemini-3-8-flash",
  "input": "Hello!",
  "instructions": "Be concise."
}
```

## Response

```json
{
  "id": "resp_example",
  "object": "response",
  "created_at": 1709884800,
  "model": "everyais/gemini-3-8-flash",
  "status": "completed",
  "output": [
    {
      "type": "message",
      "id": "msg_example",
      "status": "completed",
      "role": "assistant",
      "content": [
        {
          "type": "output_text",
          "text": "Hello! How can I help you today?",
          "annotations": []
        }
      ]
    }
  ],
  "usage": {
    "input_tokens": 25,
    "output_tokens": 10,
    "total_tokens": 35
  },
  "error": null,
  "incomplete_details": null,
  "temperature": null,
  "top_p": null
}
```

## Parameters

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `model` | string | Yes | Model ID |
| `input` | string \| array | Yes | A string, or message, `reasoning`, `function_call`, `function_call_output`, `custom_tool_call` and `custom_tool_call_output` items |
| `instructions` | string | No | System instructions |
| `temperature` | number | No | 0–2 |
| `top_p` | number | No | 0–1 |
| `stream` | boolean | No | Stream events |
| `background` | boolean | No | `true` runs an async job and returns `queued` right away |
| `tools` / `tool_choice` | array / string\|object | No | Chat form, Responses flat form, or free-form (`custom`) tools |
| `reasoning_effort` | string | No | `low`, `medium`, `high`, `xhigh`, `max`. Some models add values such as `none` or `minimal` |
| `reasoning` | object | No | OpenAI form. `effort` is the same setting as `reasoning_effort` |
| `parallel_tool_calls` | boolean | No | Same as Chat. `true` is the default and equals omitting it |
| `text` | object | No | `format` (Chat's `response_format`) and `verbosity` (`low`, `medium`, `high`; GPT models only) |
| `prompt_cache_key` / `safety_identifier` | string | No | Same as Chat. Models other than GPT drop `prompt_cache_key` with a warning |
| `user` | string | No | End-user identifier |
| `everyais` | object | No | Gateway options such as caching and `parameter_policy` |

## Differences from OpenAI

- Set reasoning effort with OpenAI's `reasoning: {"effort": "low"}` or the flat `reasoning_effort`. If you send both, they must match; different values return **400**. `reasoning.summary` is accepted and ignored.
- Stored-conversation fields such as `previous_response_id` return **400**. `store` accepts only `false`.
- `text.format` accepts OpenAI's flat `{"type":"json_schema","name","schema","strict"}` or the Chat form. Flat function tools work too.
- Free-form tools (`"type": "custom"`) work on every model. The model sees a function with one string argument, `input`, and its call comes back as a `custom_tool_call` item. The `format` grammar is passed on as a description, not enforced.
- To continue a tool conversation, send back every `output` item, including `reasoning` items, with your `function_call_output` (or `custom_tool_call_output`). Gemini's tool signature is in the `reasoning` item's `encrypted_content`.
- Images in a tool result (`input_image`) reach the model in a user message after the tool results. Models without image input return **400**.
- Web search and `response_format` are available only in Chat.

## Model options

Pick a model at the top of this page, or read `capabilities`, `limits` and `parameter_constraints` from `GET /v1/models`.
[Parameter policy](https://platform.everyais.com/en/docs/model-selection#doc-parameter-policy) decides whether unsupported values are adjusted or rejected with **400**.

Without `reasoning_effort`, `high` is used, or the model's nearest lower level when it has no `high` (on thinking-budget models, the highest of those that fits the output token limit). `none` disables reasoning only on models that list it.

---

## GET /v1/responses/{id}

Gets a job submitted with `background: true`.

```json
{
  "id": "resp_abc123",
  "object": "response",
  "created_at": 1709884800,
  "model": "everyais/gemini-3-8-flash",
  "status": "queued",
  "output": [],
  "error": null,
  "incomplete_details": null,
  "usage": null
}
```

`status` goes `queued` → `in_progress` → `completed`, `incomplete`, `failed` or `cancelled`. A finished job returns the full Response.
Missing or expired jobs, and jobs of another key or user, return 404.

## POST /v1/responses/{id}/cancel

Cancels a `queued` or `in_progress` job and refunds its credit. A finished job is returned unchanged.

---

## POST /v1/images/generations

_Generate images from a prompt. Up to 10 per request._

Source: https://platform.everyais.com/en/docs/images-generations
Updated: 2026-10-22T12:28:48.508Z

Generates images from a text prompt. Image models have `category: "IMAGE"` in `GET /v1/models`.

## Request

```json
{
  "model": "<IMAGE model ID from GET /v1/models>",
  "prompt": "a beautiful sunset over mountains",
  "aspect_ratio": "16:9",
  "resolution": "2K"
}
```

## Response

```json
{
  "created": 1709884800,
  "data": [
    { "url": "https://..." }
  ]
}
```

`url` expires after 10 minutes. Within 24 hours you can get a new link from [`GET /v1/outputs/{requestId}`](https://platform.everyais.com/en/docs/output-retrieval) using this response's `x-request-id` header, so save it.

## Parameters

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `model` | string | Yes | `IMAGE` model ID |
| `prompt` | string | Yes | 1–16,000 characters. Providers with lower limits may reject longer prompts |
| `n` | integer | No | **1–10**. Default 1 |
| `size` | string | No | `WxH` pixels (such as `256x256`, `512x512`, `1024x1024`, `1792x1024`, `1024x1792`, `1536x1024`, `1024x1536`), `auto`, or a `resolution` tier. Pixels win; a conflicting `aspect_ratio` or `resolution` returns **400** |
| `aspect_ratio` | string | No | `1:1`, `16:9`, `9:16`, `4:3`, `3:4`, `3:2`, `2:3`, `4:5`, `5:4`, `1:2`, `2:1`, `7:5`, `5:7`, `1:4`, `4:1`, `1:8`, `8:1`, `9:21`, `21:9`, `auto` (provider default) |
| `resolution` | string | No | `512`, `768`, `1K` (≈1 MP), `1.5K`, `2K` (≈4 MP), `4K` (the model's largest tier) |
| `quality` | string | No | `standard`, `hd`, `low`, `medium`, `high`, `xhigh`, `max`, `auto`. `xhigh` and `max`: GPT Image 2.5 only. See `limits.supported_qualities` |
| `response_format` | string | No | `url` (default) or `b64_json` |
| `output_format` | string | No | `png`, `jpeg`, `webp`. Only models with `output_format` in `parameter_profile`; others return **400** |
| `output_compression` | integer | No | 0–100 for `jpeg` and `webp`. GPT Image only |
| `background` | string | No | `transparent`, `opaque`, `auto`. GPT Image only. `transparent` needs `png` or `webp` |
| `moderation` | string | No | `low` or `auto`. GPT Image only |
| `user` | string | No | End-user identifier |
| `prompt_upsampling` | boolean | No | FLUX 3 Image Klein only. Default `true`. Other models return **400** |
| `extra_body.gpunode` | object | No | Open-weight models: `seed` (same seed, same image; image i of n uses seed + i) and `num_inference_steps` |
| `extra_body.google` | object | No | **deprecated**. `aspect_ratio` and `image_size` are old names for `aspect_ratio` and `resolution`. Conflicting values return **400** |

## Aspect ratio and resolution

Each model's accepted values are in `parameter_profile.tasks["image-generation"].parameters`.
Under the default `auto` policy, an unsupported ratio moves to the nearest supported one and a resolution to the nearest supported tier at or below it (or the smallest), with a note in `x_everyais.parameter_warnings`. `strict` returns **400**.

- Gemini: a pixel `size` maps to the nearest ratio and tier (`3840x2160` → `16:9`, `4K`).
- GPT Image 2.5: tiers become `WxH` on a 16 px grid (`1360x768`, `2720x1536` and `3840x2160` at 16:9). Custom sizes work with sides in multiples of 16 up to 3840, ratio 1:3–3:1 and 655,360–8,294,400 pixels.
- Models with a size list, such as FLUX 3 Image Klein, pick by ratio, then resolution. Prices by size are in `pricing.variants[].meta.sizes`.

Check `limits.supported_sizes`. Token-priced image models cost more at larger sizes and higher quality.

## FLUX 3 Image Klein

`everyais/flux-3-image-klein` is a 5.6B open-weight model from Black Forest Labs.

| Option | Value |
|---|---|
| `n` | 1–4 |
| Size | One of the 60 sizes below, or `aspect_ratio` + `resolution`. Default `1024x1024` (`1:1`, `1K`). `size: "auto"` equals omitting it |
| `output_format` | `png`, `jpeg`, `webp`. Default `jpeg` |
| `extra_body.gpunode.seed` | 0–2147483647 |
| `extra_body.gpunode.num_inference_steps` | 1–16. Default 16 |
| `prompt_upsampling` | Default `true`: BFL's upsampler rewrites the prompt into a detailed caption, as in BFL's API. `false` uses your prompt as written. `extra_body.gpunode.prompt_upsampling` is the same switch; don't send conflicting values |
| `guidance_scale`, `negative_prompt` | Not supported (**400**) |

With upsampling on, each response item includes the rewritten caption as `revised_prompt`.

### Supported sizes

Resolutions are `768`, `1K`, `1.5K` and `2K`. 4K is not supported: `auto` lowers it to `2K` and `strict` returns **400**.
Under `auto`, an unlisted `size` becomes the largest listed size of the nearest ratio that fits inside it (`1024x768` → `880x672`, `1920x1080` → `1360x768`), with a warning.
Sizes are never enlarged, so anything below `768x768` returns **400**, as does any unlisted size under `strict`. The response's top-level `size` is the actual output size.

| Aspect ratio | `768` | `1K` | `1.5K` | `2K` |
|---|---|---|---|---|
| `21:9` | 1168x496 | 1568x672 | 2352x1008 | 3136x1344 |
| `2:1` | 1088x544 | 1456x720 | 2176x1088 | 2896x1456 |
| `16:9` | 1024x576 | 1360x768 | 2048x1152 | 2736x1536 |
| `3:2` | 944x624 | 1248x832 | 1888x1248 | 2512x1680 |
| `7:5` | 912x656 | 1216x864 | 1824x1296 | 2416x1728 |
| `4:3` | 880x672 | 1184x880 | 1776x1328 | 2368x1776 |
| `5:4` | 864x688 | 1152x912 | 1712x1376 | 2288x1824 |
| `1:1` | 768x768 | 1024x1024 | 1536x1536 | 2048x2048 |
| `4:5` | 688x864 | 912x1152 | 1376x1712 | 1824x2288 |
| `3:4` | 672x880 | 880x1184 | 1328x1776 | 1776x2368 |
| `5:7` | 656x912 | 864x1216 | 1296x1824 | 1728x2416 |
| `2:3` | 624x944 | 832x1248 | 1248x1888 | 1680x2512 |
| `9:16` | 576x1024 | 768x1360 | 1152x2048 | 1536x2736 |
| `1:2` | 544x1088 | 720x1456 | 1088x2176 | 1456x2896 |
| `9:21` | 496x1168 | 672x1568 | 1008x2352 | 1344x3136 |

### Pricing (per image, USD)

| Output tier | Price |
|---|---|
| `768` | $0.008 |
| `1K` | $0.009 |
| `1.5K` | $0.011 |
| `2K` | $0.014 |

Editing costs the same. Multiply by `n`; your account rates apply.

### Bounding-box layout

Append a JSON list to the prompt to place elements: `{"id","bbox":[top,left,bottom,right],"desc"}` on a 0–1000 grid. Refer to each element in the caption as `<id>`.

```json
{
  "model": "everyais/flux-3-image-klein",
  "prompt": "A cozy cafe window scene: a <cat> sleeps on the left windowsill and a <plant> stands on the right. [{\"id\":\"cat\",\"bbox\":[500,50,900,450],\"desc\":\"orange tabby cat curled up\"},{\"id\":\"plant\",\"bbox\":[300,600,900,900],\"desc\":\"potted monstera\"}]",
  "aspect_ratio": "3:2",
  "resolution": "1K",
  "extra_body": { "gpunode": { "seed": 42 } }
}
```

### Tips

- Describe the whole scene. There is no negative prompt, so say what should be there.
- Put text you want rendered in quotes.
- License: FLUX Kommunity License (non-commercial, plus commercial use of outputs for qualifying users). Outputs pass an NSFW filter.

## Model options

Pick a model at the top of this page, or read `capabilities`, `limits` and `parameter_constraints` from `GET /v1/models`.
[Parameter policy](https://platform.everyais.com/en/docs/model-selection#doc-parameter-policy) decides whether unsupported values are adjusted or rejected with **400**. Omitted size, ratio, resolution and quality use the model's `defaults`.

---

## POST /v1/images/edits

_Edit images with a prompt and optional mask. Send multipart or JSON (base64)._

Source: https://platform.everyais.com/en/docs/images-edits
Updated: 2026-10-22T12:28:53.754Z

Edits an image from a prompt, with an optional mask or reference images. Send multipart form data, or JSON with base64 images.

## Request (JSON)

```json
{
  "model": "<ID of a model that supports editing>",
  "prompt": "change the sky to sunset",
  "image": "<base64-encoded-image>",
  "n": 1
}
```

## Response

```json
{
  "created": 1709884800,
  "data": [
    { "url": "https://..." }
  ]
}
```

## Parameters

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `model` | string | Yes | ID of a model that supports editing |
| `prompt` | string | Yes | 1–16,000 characters. Providers with lower limits may reject longer prompts |
| `image` | string \| file \| array | Yes | Base64 in JSON, a file part in multipart. FLUX 3 Image Klein takes 1–10 reference images |
| `mask` | string \| file | No | Area to edit. Needs `capabilities.image_mask: true` |
| `n` | integer | No | **1–4**. Default 1 |
| `size` | string | No | `WxH` pixels (such as `256x256`, `512x512`, `1024x1024`, `1792x1024`, `1024x1792`, `1536x1024`, `1024x1536`), `auto`, or a resolution tier. Same rules as generation |
| `aspect_ratio` | string | No | `1:1`, `16:9`, `9:16`, `4:3`, `3:4`, `3:2`, `2:3`, `4:5`, `5:4`, `1:2`, `2:1`, `7:5`, `5:7`, `1:4`, `4:1`, `1:8`, `8:1`, `9:21`, `21:9`, `auto` |
| `resolution` | string | No | `512`, `768`, `1K`, `1.5K`, `2K`, `4K` |
| `response_format` | string | No | `url` (default) or `b64_json` |
| `quality` / `output_format` / `output_compression` / `background` | | No | Same as [generation](https://platform.everyais.com/en/docs/images-generations), where supported |
| `prompt_upsampling` | boolean | No | FLUX 3 Image Klein only. `true` or `false` in multipart |
| `extra_body.gpunode` | object | No | Open-weight models: `seed`, `num_inference_steps`, `prompt_upsampling`. A JSON string in multipart |
| `extra_body.google` | object | No | **deprecated** aliases of `aspect_ratio` and `resolution`. A JSON string in multipart |
| `user` | string | No | End-user identifier |

Multipart bodies can be up to **4 MiB** and JSON up to **5 MiB**. Files are converted to base64 and must still fit in 5 MiB.
Larger requests return 413 `request_too_large` or `invalid_request_error`, so check the encoded size.

## Model options

Pick a model at the top of this page, or read `capabilities`, `limits` and `parameter_constraints` from `GET /v1/models`.
[Parameter policy](https://platform.everyais.com/en/docs/model-selection#doc-parameter-policy) decides whether unsupported values are adjusted or rejected with **400**.
Ratio and resolution rules match [image generation](https://platform.everyais.com/en/docs/images-generations#doc-aspect-ratio-and-resolution). With no size, ratio or resolution, the output usually keeps the input's aspect ratio.

## FLUX 3 Image Klein editing

For `everyais/flux-3-image-klein`:

- Send 1–10 references (JPEG, PNG, WebP). In multipart, repeat the `image` or `image[]` part. Compress them to stay within the body limits.
- Other options and prices match [generation](https://platform.everyais.com/en/docs/images-generations#doc-flux-3-image-klein).
- Without `size`, the output follows the first reference: its aspect ratio up to 4 MP and 4096 px per side, or under 1 MP on a 16 px grid when there are several references. Billing uses the nearest tier, and the response's `size` is the actual size.
- `aspect_ratio` alone gives that ratio at `1K`. `resolution` alone gives the size at that tier closest to the first reference's ratio. Explicit sizes must be one of the 60 listed sizes.
- Refer to references by position ("the first image", or `<ref_image_0>` in box edits). There is no mask.
- Describe the full target image. For box edits, append `{"id","from":"ref_image_0"|null,"src_bbox","tgt_bbox","desc"}` items (0–1000 grid; `tgt_bbox: null` removes). Boxes guide placement and scale and are not exact masks.

```json
{
  "model": "everyais/flux-3-image-klein",
  "prompt": "Move the <mug> from the first image to the right side of the table. [{\"id\":\"mug\",\"from\":\"ref_image_0\",\"src_bbox\":[600,100,800,250],\"tgt_bbox\":[600,650,800,800],\"desc\":\"white ceramic mug\"}]",
  "image": ["<base64-encoded-image>"],
  "n": 1
}
```

---

## POST /v1/videos/generations

_Generate video, billed per second. Submit a job, then poll for the result._

Source: https://platform.everyais.com/en/docs/videos-generations
Updated: 2026-10-24T22:58:02.756Z

Generates a video from a prompt and optional first or last frame. Submit a job, then poll it with the returned `id`.

## Request

```json
{
  "model": "everyais/minimax-h3-turbo",
  "prompt": "a timelapse of a blooming flower",
  "duration": 5
}
```

## Response (submitted)

```json
{
  "id": "cm...",
  "object": "video.generation.job",
  "created": 1709884800,
  "model": "everyais/minimax-h3-turbo",
  "status": "processing"
}
```

## Parameters

| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| `model` | string | Yes | `VIDEO` model ID, such as `everyais/minimax-h3-turbo` |
| `prompt` | string | Yes | 1–4000 characters |
| `duration` | integer | No | Seconds, 1–120. Supported lengths and the default vary by model (Gemini Omni 1.1: 3–10) |
| `image` | string | No | First frame: JPEG, PNG or WebP as a data URL or base64 (Veo: JPEG or PNG). Models that list `image` in `parameter_constraints` only; others return **400** |
| `last_image` | string | No | Last frame, same format. With `image`, the video goes first to last; alone, it sets only the ending (Veo and Omni require `image`). Models that list `last_image` only |
| `pricingVariant` | string | No | A `pricing.variants[].key` from `GET /v1/models`, exactly as returned |
| `user` | string | No | End-user identifier |
| `extra_body.google` | object | No | `aspectRatio` (`16:9` or `9:16`; Veo, Omni), `seed`, `enhancePrompt`, and Veo-only `negativePrompt` (up to 4000 characters) |
| `extra_body.gpunode` | object | No | Open-weight models: `size` (`WxH` from the model's list; with an image it sets only the pixel count), `seed`, `num_inference_steps` |

The request body, including images, can be up to 5 MiB.

## Model options

Pick a model at the top of this page, or read `capabilities`, `limits` and `parameter_constraints` from `GET /v1/models`.
[Parameter policy](https://platform.everyais.com/en/docs/model-selection#doc-parameter-policy) decides whether unsupported values are adjusted or rejected with **400**.
Check `limits.supported_durations` and `pricing.variants`, or omit them for the defaults. Duration and resolution must be a supported combination.

---

## GET /v1/videos/generations/{id}

Returns the job's status and result. Use the `id` from the submit response.
While the job runs, responses include `Retry-After: 5`. Poll every 5 seconds.

| `status` | Meaning |
|--------|---------|
| `processing` | In progress |
| `completed` | Done. Includes `durationSeconds` and `data` |
| `failed` | Failed or timed out. See `error` |
| `cancelled` | Cancelled |

```json
{
  "id": "cm...",
  "object": "video.generation.job",
  "created": 1709884800,
  "model": "everyais/minimax-h3-turbo",
  "status": "completed",
  "durationSeconds": 5,
  "data": [{ "url": "https://..." }]
}
```

If the result can't be confirmed yet, `processing` comes with `outcome_unknown: true`, and with `requires_manual_review: true` when it needs a manual check.
Unknown jobs, jobs older than 24 hours and other keys' jobs return 404 `job_not_found`.
`data[].url` is valid for 10 minutes, and every poll returns a fresh link. A completed video is also available from `GET /v1/outputs/{requestId}` with the job ID and the same API key (24 hours).

## Billing

Price = `durationSeconds` × per-second price. Models priced by resolution or audio list variants in `pricing.variants`; pass the `key` as `pricingVariant` (for example Veo's `720p-with-audio`).
Only models with variants accept it.

## MiniMax H3 Turbo

`everyais/minimax-h3-turbo` makes 24 fps video with stereo audio from text, a first frame, a last frame, or both.

| Option | Value |
|---|---|
| `duration` | 4–15 seconds, whole numbers. Default 5 |
| `image`, `last_image` | Both: a first-to-last transition. `last_image` alone: only the ending. Aspect ratio 0.4–2.5, short side at least 256 px |
| `extra_body.gpunode.size` | Without an image: one of the model's 13 sizes. Default `960x544`. With an image: pixel count only |
| `extra_body.gpunode.seed` | 0–2147483647 |
| `extra_body.gpunode.num_inference_steps` | 1–5. Default 5 |
| `reference_images`, `fps`, `num_frames`, `negative_prompt` | Not supported (**400**) |

With an image, the output keeps the image's aspect ratio (first frame, else last frame) at no more pixels than `size`, on a 32 px grid.
A 4:3 photo with `960x544` gives `800x608`; a 9:16 portrait with `1344x768` gives `768x1344`. The image is resized without cropping, so crop it first for a different ratio.
The per-second price is the same at every size.

```json
{
  "model": "everyais/minimax-h3-turbo",
  "prompt": "the camera slowly pushes in as she turns toward the window",
  "image": "data:image/jpeg;base64,/9j/4AAQ...",
  "duration": 5,
  "extra_body": { "gpunode": { "size": "960x544", "seed": 42 } }
}
```

Add `"last_image"` for first and last frames.

---

## GET /v1/models

_List models or get one model. Model IDs with slashes are supported._

Source: https://platform.everyais.com/en/docs/models
Updated: 2026-10-22T12:29:04.237Z

Lists the models your API key can call, with capabilities, limits and prices.

## Request

```http
GET /v1/models
```

## Response

```json
{
  "object": "list",
  "data": [
    {
      "id": "everyais/claude-sonnet-5-5",
      "aliases": ["claude-sonnet-5-5"],
      "object": "model",
      "owned_by": "everyais",
      "created": 1782008704,
      "registered_at": "2026-06-21",
      "released_at": "2026-06-21",
      "updated_at": "2026-08-27T04:30:00.000Z",
      "name": "Claude Sonnet 5.5",
      "description": "(model summary, currently Korean only)",
      "series": "claude",
      "input_modalities": ["text", "image"],
      "output_modalities": ["text"],
      "knowledge_cutoff": null,
      "price_search_per_query": null,
      "supply": "first_party",
      "training_use": "never",
      "category": "LLM",
      "type": "chat",
      "supported_endpoints": ["/v1/chat/completions", "/v1/messages", "/v1/responses"],
      "capabilities": {
        "streaming": true,
        "tool_use": true,
        "vision": true,
        "json_mode": true,
        "structured_outputs": true,
        "sampling": false,
        "parallel_tool_calls": true,
        "reasoning": true,
        "web_search": false
      },
      "limits": { "context_window": 1000000, "reasoning_efforts": ["low", "medium", "high", "xhigh", "max", "none"] },
      "parameter_profile": { "version": 1, "scope": "all-active-routes", "tasks": { "chat": { "status": "supported", "required": ["model", "messages"], "defaults": {}, "parameters": { "reasoning_effort": { "status": "supported", "type": "string", "allowed_values": ["low", "medium", "high", "xhigh", "max", "none"] } }, "combinations": [] } } },
      "parameter_constraints": [{ "provider": "aws-bedrock", "parameters": { "chat": { "reasoning_effort": { "supported": true, "type": "string", "allowed_values": ["low", "medium", "high", "xhigh", "max", "none"] } } } }],
      "pricing": {
        "unit": "per_1m_tokens",
        "is_free": false,
        "input_per_1m": 3.159,
        "output_per_1m": 15.795,
        "list": { "input_per_1m": 3.51, "output_per_1m": 17.55 },
        "discount_percent": 10
      },
      "discount_percent": 10,
      "available": true
    }
  ]
}
```

## Model fields

| Field | Description |
|-------|-------------|
| `id`, `aliases` | Model ID and other accepted names |
| `name`, `description`, `series` | Display name, description, model family |
| `category` | `LLM`, `IMAGE` or `VIDEO` |
| `supported_endpoints` | API endpoints the model works with |
| `available` | Whether the model can be called now |
| `input_modalities`, `output_modalities` | Input and output types |
| `created`, `registered_at`, `released_at` | OpenAI-compatible Unix timestamp, date added, release date |
| `knowledge_cutoff` | Knowledge cutoff, or null if unknown |
| `capabilities` | `true`: supported. `false`: not supported. Missing: not confirmed |
| `limits` | Supported values and limits |
| `parameter_profile.tasks` | Required inputs, defaults, supported values and combinations per task |
| `parameter_constraints` | Supported options per provider |

| `capabilities` key | Meaning |
|---|---|
| `streaming`, `tool_use`, `vision`, `json_mode` | Streaming, tools, image input, JSON output |
| `structured_outputs` | Strict JSON schemas |
| `json_mode_with_tools`, `web_search_with_tools` | JSON output with tools, web search with function tools |
| `web_search` | Web search in Chat |
| `sampling`, `parallel_tool_calls`, `image_mask` | Sampling options, parallel tool calls, edit masks |

`limits` includes `context_window`, `reasoning_efforts`, `supported_sizes`, `supported_qualities`, `supported_durations` and `max_images`; an empty array means nothing is supported.
Model limits can be lower than the gateway's general limits. Unsupported values return **400** `unsupported_parameter` with the field in `error.param`. See [parameter policy](https://platform.everyais.com/en/docs/model-selection#doc-parameter-policy).

## Pricing and access

| Field | Description |
|-------|-------------|
| `pricing` | Listed prices. Charges use your account rates; see the cost header and usage |
| `pricing.is_free === true` | Input, output and cache tokens are free. Web search is billed separately |
| `pricing.list`, `pricing.discount_percent` | Prices before discounts, and the smallest discount across all rates, including list-price discounts and all-customer promotions |
| `discount_percent` | The everyais promotion part, already in `pricing`. The largest value if it varies by provider |
| `pricing.variants` | Variant `key` and `price` |
| `pricing.long_context` | Prices when input exceeds `threshold_tokens`. `base_input_per_1m` and `base_output_per_1m` are aliases |
| `price_search_per_query` | Price per web search unit. See `capabilities.web_search` |
| `pricing.conditions` | Consent the listed price needs: `community_supply` or `training_use` |
| `supply`, `training_use` | Whether the model uses community supply or provider training. Off unless you opt in; organization keys follow organization settings |

Models that need a consent you haven't given are not listed.

## GET /v1/models/{model}

Gets one model by its full ID, slashes included.

```http
GET /v1/models/everyais/gemini-3-8-flash
```

Unknown or inaccessible models return 404 `model_not_found`.

---

## GET /v1/outputs/{requestId}

_Get a stored request and new output URLs, valid for 10 minutes._

Source: https://platform.everyais.com/en/docs/output-retrieval
Updated: 2026-10-24T22:58:15.337Z

Gets a completed request and new download links for its files. Requests are kept for **24 hours**.
Use an API key or a signed-in session; you can retrieve only your own requests.

## Find the requestId

Use the `x-request-id` header from an image generation or edit response (`x-everyais-operation-id` has the same value).
It isn't in the response body, so save the header. An `x-request-id` you send yourself is only for tracing and doesn't become the ID.

```http
HTTP/1.1 200 OK
content-type: application/json
x-request-id: V1StGXR8_Z5jdHi6B-myT
x-everyais-operation-id: V1StGXR8_Z5jdHi6B-myT
```

## Retrieve the result

```http
GET /v1/outputs/V1StGXR8_Z5jdHi6B-myT
```

`urls` are presigned links valid for 10 minutes.

```json
{
  "id": "V1StGXR8_Z5jdHi6B-myT",
  "model": "everyais/gemini-3-1-flash-image",
  "endpoint": "/v1/images/generations",
  "status": "COMPLETED",
  "inputTokens": 0,
  "outputTokens": 0,
  "imageCount": 1,
  "totalCost": 0.04,
  "latencyMs": 1200,
  "urls": ["https://..."],
  "createdAt": "2026-07-30T01:23:45.000Z"
}
```

Requests outside your key's scopes return 403 `insufficient_scope`. For a completed video, use the job ID (the submit response's `id`) as `requestId`. Poll progress with `GET /v1/videos/generations/{id}`.

---

# Guides

## Streaming

_Show the answer as it is generated instead of waiting for the full response._

Source: https://platform.everyais.com/en/docs/streaming
Updated: 2026-10-22T12:30:26.550Z

Streaming sends the response as it is generated, so you can show the first tokens right away. Use it where a person is waiting, such as a chat screen.

## Try it

Use the environment variables from the [quickstart](https://platform.everyais.com/en/docs/quickstart). Add `stream: true`, plus `stream_options.include_usage: true` for token usage at the end.

### Python

```python
import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["EVERYAIS_API_KEY"],
    base_url="https://api.everyais.com/v1",
    timeout=240.0,
)
stream = client.chat.completions.create(
    model=os.environ["EVERYAIS_MODEL"],
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True,
    stream_options={"include_usage": True},
)
try:
    for chunk in stream:
        if chunk.choices:
            print(chunk.choices[0].delta.content or "", end="", flush=True)
        if chunk.usage:
            print("\nUsage:", chunk.usage.total_tokens)
finally:
    stream.close()
```

### Node.js

```javascript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.everyais.com/v1",
  apiKey: process.env.EVERYAIS_API_KEY,
  timeout: 240_000,
});
const stream = await client.chat.completions.create({
  model: process.env.EVERYAIS_MODEL,
  messages: [{ role: "user", content: "Hello!" }],
  stream: true,
  stream_options: { include_usage: true },
});
for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
  if (chunk.usage) console.log("\nUsage:", chunk.usage.total_tokens);
}
```

## Handle these cases

- The final usage chunk has an empty `choices` array. Check it before you read `choices[0]`.
- A Chat stream ends with `data: [DONE]`. If the connection closes without it, the response was cut off.
- An error can arrive as an event after the stream started with HTTP 200. Treat error events and SDK exceptions as failures.
- To stop generation, close the stream (`stream.close()` in Python, an `AbortController` in Node.js). The provider may not stop right away, and those tokens can still be billed.

## Parsing SSE yourself

- Events are `data: {...}` lines, and a blank line ends each event. Buffer network chunks until you reach one.
- Lines starting with a colon, such as `: ok`, are keep-alive comments. Skip them.
- Decode UTF-8 with a streaming decoder, since characters can be split across chunks.

## Messages and Responses streams

These use different events from Chat. Use each SDK's stream helpers instead of Chat `[DONE]` handling.

| API | Main events |
|-----|-------------|
| `/v1/messages` | `message_start`, `content_block_delta`, `message_delta`, `message_stop` |
| `/v1/responses` | `response.output_text.delta`, `response.completed` |

With `background: true`, Responses returns a job instead of a stream. See [async jobs](https://platform.everyais.com/en/docs/async-jobs).

## Retrying a stream

`Idempotency-Key` can't replay a finished stream; resending the same key returns 409. Before you retry, check the output you already received and your usage.
With web search, citations and the search count arrive at the end. See [web search](https://platform.everyais.com/en/docs/web-search#doc-streaming).

---

## Prompt caching

_Cache the prefix you send on every request to cut input cost. Check the math first._

Source: https://platform.everyais.com/en/docs/prompt-caching
Updated: 2026-10-24T22:58:32.113Z

Caching helps when every request starts with the same long prefix, such as a system prompt or tool definitions. Cached prefixes are read at about 10% of the input price.
If the prefix changes on every request, leave caching off: cache writes cost more than normal input.

## Turn it on

```json
{
  "model": "everyais/claude-sonnet-5-5",
  "messages": [
    {"role": "system", "content": "(the long instructions you send every time)"},
    {"role": "user", "content": "First question"}
  ],
  "everyais": { "cache": "on", "cache_ttl": "5m" }
}
```

With the OpenAI SDK, pass it in `extra_body`.

```python
response = client.chat.completions.create(
    model="everyais/claude-sonnet-5-5",
    messages=messages,
    extra_body={"everyais": {"cache": "on", "cache_ttl": "5m"}},
)
print(response.usage.prompt_tokens_details.cached_tokens)
```

| Option | Behavior |
|------|------|
| `cache: "on"` | Adds up to 4 cache points: tool definitions, system prompt, then the last user turn |
| `cache: "off"` | Removes automatic cache points and any `cache_control` you sent |
| `cache_ttl` | `"5m"` (default) or `"1h"`. One-hour writes cost more |

Gemini models cache automatically, with no setup and no write charge. On `/v1/messages`, `cache_control` on blocks is passed through unchanged.

## Cost example

A cache read costs about 10% of the input price. A write costs about 1.25 times input for 5 minutes, or 2 times for 1 hour.
For a 10,000-token prefix sent five times within five minutes, in input-token equivalents:

| Case | Input-token equivalent |
|------|-----------------|
| Caching off | 10,000 × 5 = 50,000 |
| Caching on, same prefix | 12,500 (first-request write) + 1,000 × 4 = 16,500 |
| Caching on, prefix changes every time | 12,500 × 5 = 62,500 |

Messages after the prefix are billed at the normal input price.

## Check for cache hits

`usage.prompt_tokens_details.cached_tokens` counts tokens read from cache, and `cache_creation_tokens` counts tokens written, when the provider reports them. Both are billed at the provider's cache prices.
`prompt_tokens` (`input_tokens` on Responses) is the total input including both. On `/v1/messages`, `input_tokens` excludes cache reads and writes, as in Anthropic's API.

If `cached_tokens` stays at 0:

- The prefix must match exactly. Keep changing values, such as dates or request IDs, out of the start of the system prompt.
- The cache expires after `cache_ttl`.
- Prefixes shorter than the model's minimum are not cached.

---

## Async jobs

_Submit long-running work, then fetch the result with the job ID._

Source: https://platform.everyais.com/en/docs/async-jobs
Updated: 2026-10-24T22:58:39.456Z

For work that takes too long for one HTTP request, submit a job, get an ID back, and poll until it finishes.

| Task | Submit | Get the result |
|-----------|------|-----------|
| Long text generation | `POST /v1/responses` with `background: true` | `GET /v1/responses/{id}` |
| Video generation | `POST /v1/videos/generations` (always async) | `GET /v1/videos/generations/{id}` |

The IDs are not interchangeable. A video job ID returns 404 from a Responses endpoint. A completed video is also available from `GET /v1/outputs/{requestId}` with the job ID.

## Responses background jobs

With `background: true`, the API returns `status: "queued"` right away. Create the client as in the [quickstart](https://platform.everyais.com/en/docs/quickstart).

```python
import time

job = client.responses.create(
    model="everyais/gemini-3-8-flash",
    input="A long research task...",
    background=True,
)
while job.status in ("queued", "in_progress"):
    time.sleep(5)
    job = client.responses.retrieve(job.id)

print(job.status)
print(job.output_text)
```

The job ends as `completed`, `incomplete`, `failed` or `cancelled`. Cancel a `queued` or `in_progress` job with `POST /v1/responses/{id}/cancel` (`client.responses.cancel(id)`); its credit is refunded.

## Video jobs

Poll with the `id` from the submit response. Responses include `Retry-After: 5` while the job runs.

```python
import os
import time
import requests

BASE_URL = "https://api.everyais.com/v1"
headers = {"Authorization": f"Bearer {os.environ['EVERYAIS_API_KEY']}"}

job = requests.post(f"{BASE_URL}/videos/generations", headers=headers, json={
    "model": "everyais/minimax-h3-turbo",
    "prompt": "a timelapse of a blooming flower",
}).json()

while job["status"] == "processing":
    time.sleep(5)
    job = requests.get(f"{BASE_URL}/videos/generations/{job['id']}", headers=headers).json()

print(job["status"], job.get("data"))
```

The job ends as `completed`, `failed` or `cancelled`, and its record is kept for 24 hours. See [video generation](https://platform.everyais.com/en/docs/videos-generations#doc-get-v1videosgenerationsid) for status details.

## Avoid duplicate charges

- Don't resubmit a video because you missed the submit response. Video submission doesn't support `Idempotency-Key`, so a resubmit creates and bills a second job.
- Polling too often uses your key's rate limit. Follow `Retry-After`.

---

## Web search

_Let the model search the web for current information. Covers search billing and citations._

Source: https://platform.everyais.com/en/docs/web-search
Updated: 2026-10-25T17:27:49.330Z

Add `{"type": "web_search"}` to `tools`. The model searches when it needs to and answers from the results. The search runs server-side, so your client runs no tool.

```json
{
  "model": "everyais/gemini-3-8-flash",
  "messages": [{"role": "user", "content": "What's the weather in Seoul today?"}],
  "tools": [{"type": "web_search"}]
}
```

## Supported models

Use models with `capabilities.web_search: true` in `GET /v1/models`. Support varies even within the Gemini family.

```bash
curl https://api.everyais.com/v1/models \
  -H "Authorization: Bearer $EVERYAIS_API_KEY" \
  | jq '.data[] | select(.capabilities.web_search) | .id'
```

Other models return 400 `web_search_unsupported_model` and are not charged.

- Only one `web_search` entry is allowed. Two or more return 400.
- You can combine it with function tools. `tool_choice: "none"` turns off both. With only the search tool, `required` or a named function returns 400.
- Only `/v1/chat/completions` supports web search.

## Billing

Search is billed on top of tokens: `billed_queries` × the model's search price.

| Model family | How searches are counted |
|-----------|-----------|
| Gemini | Each query run. Two searches count as 2; no search counts as 0 |

- Search is billed even when tokens are free. `pricing.is_free` covers only input, output and cache tokens.
- A stream you cut off still pays for the searches Google ran. Google reports the count at the end of the answer, so the gateway waits up to 10 seconds after the cut for it; if it doesn't arrive, the 10 searches reserved with the request are charged.
- Pages fetched by Google search are not billed as input tokens.
- `x_everyais.web_search.billed_queries` has the billed count. The `x-everyais-cost-usd` header on non-streaming responses includes search.
- You can't cap the number of searches. Use the key's monthly or daily spend limit instead.

## Citations

```json
{
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "Seoul is clear today with a high of 28 degrees.",
      "annotations": [
        {
          "type": "url_citation",
          "url_citation": {
            "url": "https://...",
            "title": "Seoul weather",
            "start_index": 0,
            "end_index": 47
          }
        }
      ]
    },
    "finish_reason": "stop"
  }],
  "x_everyais": {
    "web_search": {
      "queries": ["Seoul weather today"],
      "search_entry_point_html": "<div>...</div>",
      "billed_queries": 1
    }
  }
}
```

| Field | Description |
|------|------|
| `message.annotations[]` | OpenAI `url_citation` citations. `start_index` and `end_index` are character positions in `content` |
| `x_everyais.web_search.queries` | Queries the model ran |
| `x_everyais.web_search.billed_queries` | Searches billed |
| `x_everyais.web_search.search_entry_point_html` | Google search suggestions as HTML |

If you show search results, Google requires you to display `search_entry_point_html`. Render this third-party HTML isolated, for example in `<iframe sandbox srcdoc="...">`.

## Streaming

With `stream: true`, events arrive in this order:

1. `delta.content` chunks
2. One `delta.annotations` chunk
3. The `finish_reason` chunk
4. The usage chunk (`choices: []`) with `x_everyais.web_search`

Read the stream to the end if you track citations or cost.

---

# Coding tools

## Claude Code / Anthropic SDK

_Connect Claude Code and the Anthropic SDK to everyais, with a settings.json built from the models you pick._

Source: https://platform.everyais.com/en/docs/integration-claude-code
Updated: 2026-10-22T12:29:30.455Z

Claude Code sends requests in the Anthropic Messages format (`/v1/messages`). Change the base URL and the key, and it works with everyais.

## Connect

1. Create an API key in the [dashboard](https://platform.everyais.com/dashboard/keys) and add credits.
2. In "Generate your config" above, pick your models and press Confirm. Only models that support tool calling are listed. Claude models also get a "Reasoning effort" choice.
3. Paste the result into `~/.claude/settings.json` and replace `everyais_your_api_key` with your key. If the file already exists, merge in the same keys.
4. Restart Claude Code and check that your models are listed in `/model`.

Keep the key in your user settings file (`~/.claude/settings.json`). A project's `.claude/settings.json` gets committed to the repository, which would publish the key.

## Model order

Your picks fill Claude Code's model aliases in order.

| Pick | Slot |
|------|------|
| 1st | The default model and `opus` |
| 2nd | `sonnet` |
| 3rd | `haiku`, which also runs background work |

With fewer than three picks, earlier models fill the empty slots. Pick a fast, inexpensive model last.

## Settings

| Key | What it does |
|-----|--------------|
| `ANTHROPIC_BASE_URL` | Sends requests to everyais. Claude Code appends `/v1/messages` itself, so leave `/v1` off |
| `ANTHROPIC_AUTH_TOKEN` | Your everyais key, sent as an `Authorization: Bearer` header and used ahead of a signed-in claude.ai account |
| `ANTHROPIC_DEFAULT_OPUS_MODEL` · `ANTHROPIC_DEFAULT_SONNET_MODEL` · `ANTHROPIC_DEFAULT_HAIKU_MODEL` | The three slots above. If they are empty, Claude Code may request its own default model names and get 404 |
| `CLAUDE_CODE_MAX_OUTPUT_TOKENS` | Maximum output tokens per request. It applies to every model, so it is the smallest output limit among your picks. It is left out if any pick's limit is unknown |
| `CLAUDE_CODE_MAX_CONTEXT_TOKENS` | Context size for models that are not Claude. Without it, Claude Code assumes 200K tokens and compacts the conversation at that size. It does not apply to Claude models |
| `CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS` | Turns off request fields for experimental features. everyais does not accept some of them yet |
| `model` | The default model when Claude Code starts |
| `modelPicker` | The `/model` list. `replaceBuiltInOptions: true` shows your picks instead of Claude Code's built-in list. Applies only from the user settings file and needs Claude Code 2.1.242 or later |
| `modelSettings.<model>.effortLevel` | Reasoning effort for a Claude model: `low`, `medium`, `high`, or `xhigh`. This key does not accept `max`. Inside Claude Code, change it with `/effort` |

You can also pass the key as `ANTHROPIC_API_KEY`. Claude Code then sends it as an `x-api-key` header and asks once whether to use it.

Note: Claude Code counts tokens with `POST /v1/messages/count_tokens` before each request. That call is free. Usage and spend are on the [dashboard usage page](https://platform.everyais.com/dashboard/usage).

## Models that are not Claude

You can pick models such as Gemini or GPT. Claude Code builds its requests for Claude, so tool calls and the thinking display may look different. Confirm the connection with a Claude model first.
Claude Code sends no reasoning effort to these models, so they answer at their default effort.

## Anthropic SDK

The official Anthropic SDKs (Python and TypeScript) read these two environment variables, so your code stays the same.

```bash
export ANTHROPIC_BASE_URL="https://api.everyais.com"
export ANTHROPIC_API_KEY="everyais_your_api_key"
```

Use everyais model IDs such as `everyais/claude-sonnet-5-5`. Supported parameters are listed in [Messages](https://platform.everyais.com/en/docs/messages).

## Troubleshooting

| Symptom | What to check |
|---------|---------------|
| 404 with `/v1/v1` in the URL | Remove the trailing `/v1` from the base URL |
| 401 | Check the key again, including stray spaces around it |
| 404 `model_not_found` | Check the model IDs and the key's allowed models. All three `ANTHROPIC_DEFAULT_*_MODEL` values must be set |
| 400 | Check that `CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS` is `"1"`. If it persists, look at the field named in `error.param` |
| Requests still go to your claude.ai subscription | Save settings.json and fully restart Claude Code |

---

## Cursor

_Change the OpenAI base URL to use everyais models in Cursor chat and agent._

Source: https://platform.everyais.com/en/docs/integration-cursor
Updated: 2026-10-23T14:42:47.650Z

Put your everyais key and base URL into Cursor's OpenAI API key settings. Requests go through Cursor's servers to everyais. Cursor does not fetch a model list from the base URL, so you add the models you want yourself.

## Connect

1. In Cursor, open Settings (⚙️) → **Models** and expand **API Keys** below the model list.
2. Turn on **Override OpenAI Base URL** and enter `https://api.everyais.com/v1` in **Base URL**. Do this before the key: Cursor sends its key check to this URL.
3. Enter your everyais key in the OpenAI **API Key** field, turn on **Use OpenAI API Key**, and confirm "Enable OpenAI API Key". Cursor sends a check request to that URL and the switch stays on once it passes (older versions show a **Verify** button instead).
4. In **Add or search model** above the model list, enter each model ID you picked in "Generate your config", exactly as shown including `everyais/`, and press **Add Custom Model** → **Add**. Repeat per model and make sure its switch is on.
5. In the chat input's model menu, pick an everyais model you added.

## Good to know

- Models Cursor already lists are not everyais models. Names starting with `claude-` or `gemini-` (such as `claude-opus-5-5`) run on Cursor's own billing. Other models (such as GPT) are sent to everyais while the base URL override is on, and fail with `model_not_found` if everyais has no model by that name.
- This applies to chat and agent models. Tab autocomplete keeps using Cursor's own model.
- Turn the OpenAI API Key switch off to go back.
- Cursor may hide thinking text that comes from a custom API.
- Cursor has no per-model fields for context, output limits, or prices. "Generate your config" gives only the values the connection needs.
- Cursor shows no reasoning effort setting for added models, and with a custom base URL it doesn't send one. Add the model with the effort at the end of its ID instead (for example `everyais/gemini-3-8-flash:high`). The value is one of `none`, `minimal`, `low`, `medium`, `high`, `xhigh`, or `max` that the model accepts. To switch effort, add the same model again with another level. A value the model doesn't accept is changed to one it does, with a note in `x_everyais.parameter_warnings`.

## Troubleshooting

| Symptom | What to check |
|---------|---------------|
| The Use OpenAI API Key switch will not stay on | Check that you set the base URL before the key, that it ends in `/v1`, and that the key has no spaces |
| 404 `model_not_found` | Check that you picked an everyais model you added, that its ID includes `everyais/`, and that the key allows it |
| A model name error in agent mode | Some Cursor plans and versions block custom models there. Try chat mode first |

---

## opencode

_Register every everyais model in opencode with one command, or set it up yourself in opencode.json._

Source: https://platform.everyais.com/en/docs/integration-opencode
Updated: 2026-10-22T17:02:56.834Z

Add everyais to opencode as an OpenAI-compatible provider. One command puts every everyais model in `/models`.

## Automatic setup (recommended)

1. Run this in a terminal.

   ```bash
   opencode auth login https://platform.everyais.com
   ```

2. When `everyais API key:` appears, paste the key from your dashboard and press Enter. The key is not shown as you type.
3. Start opencode and pick an everyais model in `/models`.

Every everyais model that supports tool calling is registered with its context, output limit, and prices. opencode fetches this setup each time it starts, so new models and price changes show up on their own. For models that support reasoning, change the effort with `Ctrl+T` or `opencode run --variant high`.

- Works on macOS, Linux, and WSL (needs `sh`). On Windows, use the file setup below.
- To disconnect, run `opencode auth logout` and choose `https://platform.everyais.com`.

## Set it up in a file

To pick models yourself or set a default model, write them into opencode.json.

1. Put the key in an environment variable. Add it to your shell profile (`~/.zshrc` and so on) so you don't have to type it each time.

   ```bash
   export EVERYAIS_API_KEY="everyais_your_api_key"
   ```

2. In "Generate your config" above, pick your models and press Confirm. Only models that support tool calling are listed.
   The 1st model becomes the default (`model`) and the 2nd handles light work such as title generation (`small_model`). Models that support reasoning also get a "Reasoning effort" choice.
3. Save the result as `~/.config/opencode/opencode.json` to use it in every project, or as `opencode.json` at a project root to use it in that project only.
   If the file already exists, merge only the `everyais` entry under `provider` plus `model` and `small_model`.
4. Start opencode and check that everyais models show up in `/models`.

## Settings

| Key | What it does |
|-----|--------------|
| `npm: "@ai-sdk/openai-compatible"` | The adapter that talks to OpenAI-compatible APIs |
| `options.baseURL` | `https://api.everyais.com/v1`, including `/v1` |
| `options.apiKey` | `{env:EVERYAIS_API_KEY}` reads the key from the environment. If the variable is empty, requests get 401 |
| `model` | The default model. opencode reads everything before the first `/` as the provider name, so it is `everyais/` followed by the model ID, with `everyais` appearing twice |
| `small_model` | Your 2nd pick, used for light work such as title generation |

Each model gets these values from the catalog. Values the catalog does not have are left out.

| Key | Value |
|-----|-------|
| `name` | Display name |
| `tool_call` | Whether the model supports tool calling |
| `modalities.input` | Accepted input. Models that accept images get `image` |
| `limit.context` · `limit.output` | Context size and maximum output tokens. opencode needs both together, so they are set only when both are known |
| `cost` | USD per 1M tokens for input, output, cache read, and cache write, plus the price for input over 200K tokens (`context_over_200k`) |
| `options.reasoningEffort` | The reasoning effort you chose |

To store the key in opencode instead, run `/connect`, choose Other, enter `everyais` as the provider ID, and paste the key. You can then delete the `options.apiKey` line.

## Example

```json
{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "everyais": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "everyais",
      "options": {
        "baseURL": "https://api.everyais.com/v1",
        "apiKey": "{env:EVERYAIS_API_KEY}"
      },
      "models": {
        "everyais/claude-sonnet-5-5": { "name": "Claude Sonnet 5.5", "tool_call": true },
        "everyais/gemini-3-8-flash": { "name": "Gemini 3.8 Flash", "tool_call": true }
      }
    }
  },
  "model": "everyais/everyais/claude-sonnet-5-5",
  "small_model": "everyais/everyais/gemini-3-8-flash"
}
```

## Troubleshooting

| Symptom | What to check |
|---------|---------------|
| `failed to fetch remote config` on startup | Automatic setup fetches its config from `https://platform.everyais.com` each time opencode starts. Check your network, or disconnect with `opencode auth logout` |
| 401 | With automatic setup, run `opencode auth login` again and enter the key. With the file setup, run `echo $EVERYAIS_API_KEY` in the shell that starts opencode to check the key is set |
| 404 `model_not_found` | Check that each key under `models` exactly matches an everyais model ID |
| The default model is not found | Check that `model` has the `everyais/everyais/...` form |

---

## Continue

_Add everyais models to Continue's config.yaml, with every model you pick registered._

Source: https://platform.everyais.com/en/docs/integration-continue
Updated: 2026-10-25T17:44:17.001Z

Continue (the VS Code and JetBrains extension) registers models in `config.yaml` and lets you pick one in the chat panel.

## Connect

1. Create `~/.continue/.env` and put the key in it, without quotes.

   ```bash
   EVERYAIS_API_KEY=everyais_your_api_key
   ```

2. In "Generate your config" above, pick your models and press Confirm. Models for Agent mode must support tool calling, and those models get `tool_use` in `capabilities`. Models that support reasoning also get a "Reasoning effort" choice.
3. Paste the result into `~/.continue/config.yaml`. If you already have models, add only the entries under `models:`.
4. Pick an everyais model from the model list in Continue's chat panel.

The VS Code and JetBrains extensions cannot read shell environment variables. A key set with `export` will not be seen, so keep it in a `.env` file.

## Settings

| Key | What it does |
|-----|--------------|
| `provider: openai` · `apiBase` | Sends requests to everyais through the OpenAI-compatible API. `apiBase` includes `/v1` |
| `apiKey: ${{ secrets.EVERYAIS_API_KEY }}` | Reads the key from a `.env` file: the workspace `.env`, then `.continue/.env`, then `~/.continue/.env` |
| `roles` | Uses the model for chat (`chat`), code edits (`edit`), and applying changes (`apply`). The autocomplete (`autocomplete`) role is left out until it has been tested |
| `capabilities` | Tool calls (`tool_use`, needed for Agent mode) and image input (`image_input`). Set only on models that support them |
| `defaultCompletionOptions.contextLength` · `maxTokens` | Context size and maximum output tokens. Left out when the catalog has no value |
| `requestOptions.extraBodyProperties.reasoning_effort` | The reasoning effort you chose, sent as-is in the request body |
| `useResponsesApi: false` | Keeps requests on Chat Completions. Added only for models whose name contains `gpt-5` or later, such as `gpt-6` |

## Example

```yaml
name: everyais
version: 1.0.0
schema: v1
models:
  - name: "Claude Sonnet 5.5"
    provider: openai
    model: "everyais/claude-sonnet-5-5"
    apiBase: https://api.everyais.com/v1
    apiKey: ${{ secrets.EVERYAIS_API_KEY }}
    roles: [chat, edit, apply]
    capabilities: [tool_use, image_input]
```

## Troubleshooting

| Symptom | What to check |
|---------|---------------|
| 401 | Check that `~/.continue/.env` has the key and the variable is named `EVERYAIS_API_KEY` |
| The model cannot be picked in Agent mode | Check that the model has `capabilities: [tool_use]` |
| 400 on a GPT model | Continue sends these models to `/v1/responses`. Check that the model entry has the `useResponsesApi: false` line the generator adds |
| 404 `model_not_found` | Check that `model` matches an everyais model ID |

---

## Cline

_Connect everyais through Cline's OpenAI Compatible provider. Plan and Act can use different models._

Source: https://platform.everyais.com/en/docs/integration-cline
Updated: 2026-10-24T22:58:25.823Z

Cline is set up from its settings screen. Pick models in "Generate your config" above to get the values to paste.
Cline edits files through tool calls, so only models that support tool calling are listed.

## Connect

1. Open settings (⚙️) in the Cline panel.
2. Under **API Provider**, choose **OpenAI Compatible**.
3. Enter `https://api.everyais.com/v1` as the **Base URL** and your everyais key as the **API Key**.
4. Enter the model ID in **Model ID**. If Cline loads the everyais model list, you can pick from it instead.
5. Expand **MODEL CONFIGURATION** under Model ID and enter **Supports Images**, **Context Window Size**, **Max Output Tokens**, **Input Price / 1M tokens**, and **Output Price / 1M tokens** as shown above. Leave any field that is not shown empty, because the catalog has no value for it.
6. For models that support reasoning, set **Reasoning Effort** to the value shown.

## Different models for Plan and Act

Cline uses one model per mode. Turn on **Use different models for Plan and Act modes** to set each mode separately.
If you pick two or more models above, the 1st is shown for Plan Mode and the 2nd for Act Mode. Pick the stronger model first, for planning.
Each mode gets its own values, and later picks are listed as model IDs to switch to.

## Troubleshooting

| Symptom | What to check |
|---------|---------------|
| 401 | Paste the key into the API Key field again |
| 404 | Check that the base URL ends in `/v1` and the Model ID matches an everyais model ID |
| Tool calls do not work | Check in the [model directory](https://platform.everyais.com/en/models) that the model supports tool calling |

---

## Other OpenAI-compatible tools

_Connect Zed, Aider, and tools built on the OpenAI SDK._

Source: https://platform.everyais.com/en/docs/integration-openai-compatible-tools
Updated: 2026-10-22T12:29:56.664Z

Most tools that use the OpenAI format work once you set the provider to OpenAI (or OpenAI Compatible) and enter these values.

| Field | Value |
|-------|-------|
| Base URL | `https://api.everyais.com/v1` (including `/v1`) |
| API Key | `everyais_...` |
| Model | A model ID you picked in "Generate your config" above (e.g. `everyais/gemini-3-8-flash`) |

## Tools built on the OpenAI SDK

The official OpenAI SDKs (Python and Node.js) read the `OPENAI_BASE_URL` and `OPENAI_API_KEY` environment variables.
A tool that uses the SDK as-is sends requests to everyais once you set the two `.env` lines generated above.

## Zed

Merge the generated `language_models` block into the existing object in `~/.config/zed/settings.json`. If `openai_compatible` already exists, add only `everyais` under it.

Each model gets its context size, maximum output tokens, reasoning effort (`reasoning_effort`), and supported features (`capabilities`) from the catalog.
Zed requires a context value, so models with no context in the catalog are left out.

Provide the key through the `EVERYAIS_API_KEY` environment variable or Zed's settings screen, not in settings.json. Pick a registered model from the agent panel's model list.

## Aider

Aider sends a model to an OpenAI-compatible API when its name starts with `openai/`, so put `openai/` in front of the everyais model ID, as in `openai/everyais/gemini-3-8-flash`.
Put the generated files in your home folder or the repository root, and pass the key in the `OPENAI_API_KEY` environment variable.

| File | Contents |
|------|----------|
| `.aider.conf.yml` | The main model, the `weak-model` for light work such as commit messages, and model aliases |
| `.aider.model.metadata.json` | Context and output limits, per-token prices, and tool, image, and reasoning support |
| `.aider.model.settings.yml` | The reasoning effort sent for each model. It appears only when you pick a model that accepts reasoning effort |

Per-token prices in the metadata are the per-1M price divided by 1,000,000 (for example, $2 → 0.000002).
The 1st model becomes the main model and the 2nd becomes the `weak-model`.
In chat, switch models with `/model` followed by an alias, and change reasoning effort with `/reasoning-effort`.

## Model names

Short names without the `everyais/` prefix, such as `claude-sonnet-5-5`, also work, but use full IDs in config files. The rules are in [compatibility features](https://platform.everyais.com/en/docs/compatibility-features#doc-shorter-model-names).

---

# Reference

## Errors and retries

_Error format, what to do for each status code, rate limits, and safe retries._

Source: https://platform.everyais.com/en/docs/errors-and-rate-limits
Updated: 2026-10-25T04:30:22.562Z

## Error format

Inference APIs return errors in the OpenAI format. `/v1/messages` can use the Anthropic format (`type: "error"`).

```json
{
  "error": {
    "message": "Incorrect API key provided. ...",
    "type": "authentication_error",
    "param": null,
    "code": "invalid_api_key"
  }
}
```

Branch on `error.code`; `message` wording can change. `param` names the field that caused the error.
Errors after a stream starts arrive as stream events. See [streaming](https://platform.everyais.com/en/docs/streaming). The event's `error.code` is the same as in the table below (for example `provider_unavailable` or `upstream_timeout`). Stream events carry no `Retry-After`, so wait a few seconds before resending.

## What to do for each status

| Status | type / code | Cause | Retry? |
|------|-------------|---------------|--------------------|
| 400 | `invalid_request_error` / `unsupported_parameter` · `unsupported_value` · `image_input_unsupported` · `context_length_exceeded` | Invalid parameter, context too long, or an option or image input the model doesn't support | No. Fix the request |
| 401 | `authentication_error` / `missing_api_key` · `invalid_api_key` | Missing, invalid or expired key | No |
| 402 | `insufficient_quota_error` / `insufficient_quota` · `spend_limit_exceeded` | Not enough credit, or the key's spend limit | After adding credit or raising the limit |
| 403 | `permission_error` / `model_not_allowed` · `insufficient_scope` | Model or scope not allowed for this key | No |
| 404 | `invalid_request_error` / `model_not_found` · `job_not_found` | Wrong model ID, model blocked by your settings, or expired job | No |
| 409 | `idempotency_*` | `Idempotency-Key` collision. See below | Depends |
| 413 | `invalid_request_error` / `request_too_large` | Body over the [size limit](https://platform.everyais.com/en/docs/base-url#doc-request-body-size) | After making it smaller |
| 429 | `rate_limit_error` / `rate_limit_exceeded` | Per-minute request limit exceeded | Yes, after `Retry-After` |
| 502 | `api_error` / `backend_unavailable` · `upstream_error` | The model provider returned an error | Yes, with backoff |
| 503 | `api_error` / `model_warming` | The model is starting | Yes, after `Retry-After` |
| 503 | `api_error` / `provider_unavailable` | Every route for this model is briefly blocked. Your request was not run | Yes, after `Retry-After` (30 s at most) |
| 504 | `api_error` / `upstream_timeout` | The model provider did not respond in time | Yes, with backoff |

When a provider rejects the request, the 400 carries the provider's field and code in `param` and `code` (for example `param: "reasoning_effort"`, `code: "unsupported_value"`). The field is renamed to the API you called: a Chat Completions request gets `messages`, not the provider wire's `input`. The provider's own message text is not included.

For 404 `model_not_found` on an ID that looks right, check the key's allowed models and your [access settings](https://platform.everyais.com/en/docs/model-selection#doc-3-check-access-policies).

### 409: when an Idempotency-Key collides

| code | Meaning | What to do |
|------|----|-------|
| `idempotency_in_progress` | The same request is still running | Wait. Don't resend with a new key |
| `idempotency_key_reused` | A different body was sent with the same key | Use a new key per distinct request |
| `idempotency_response_not_replayable` | It already ran, but a stream or large result can't be replayed | Check the output and your usage |

### 503 model_warming

The model is starting. Your request was not run. Retry after `Retry-After` seconds. Some models take 10–20 minutes to start.

## Rate limits

Each key allows **100 requests per minute** by default; change it when you create or edit the key. Login, sign-up and other auth endpoints have a lower limit.

| Header | Meaning |
|------|----|
| `X-RateLimit-Limit` | The key's per-minute limit |
| `X-RateLimit-Remaining` | Requests left in the window |
| `X-RateLimit-Reset` | Reset time (Unix milliseconds) |
| `Retry-After` | Seconds to wait before retrying a 429 or 503 |

## Retry code

The OpenAI SDK retries 429 and 5xx twice with short waits. A 503 `model_warming` can last minutes, so retry yourself.

```python
import os
import random
import time
from openai import APIStatusError, OpenAI

client = OpenAI(
    base_url="https://api.everyais.com/v1",
    api_key=os.environ["EVERYAIS_API_KEY"],
    max_retries=0,  # we retry below
)
RETRYABLE = {429, 502, 503, 504}

def create_with_retry(attempts=5, **request):
    for attempt in range(attempts):
        try:
            return client.chat.completions.create(**request)
        except APIStatusError as error:
            if error.status_code not in RETRYABLE or attempt == attempts - 1:
                raise
            wait = float(error.response.headers.get("retry-after") or 2 ** attempt)
            time.sleep(wait + random.random())
```

- Send an `Idempotency-Key` with automatic retries so a request doesn't run twice. See [compatibility features](https://platform.everyais.com/en/docs/compatibility-features#doc-retry-the-same-request-safely).
- Video submission doesn't support `Idempotency-Key`. If you lose the submit response, check the [job status](https://platform.everyais.com/en/docs/videos-generations) before resubmitting.

## Contacting support

Include the `x-request-id` response header, the HTTP status and `error.code`.

---

## Compatibility features

_Shorter model names, per-request cost, safe retries, and per-endpoint limits._

Source: https://platform.everyais.com/en/docs/compatibility-features
Updated: 2026-10-22T12:30:37.560Z

## Shorter model names

You can drop the `everyais/` prefix: `claude-sonnet-5-5` resolves to `everyais/claude-sonnet-5-5`, and the dotted `gemini-3.8-flash` to `everyais/gemini-3-8-flash`.
Registered provider model names also work. Accepted names are in `aliases` in `GET /v1/models`.

A short name works only if it matches exactly one available model; otherwise the API returns 404. Use full IDs in config and code, since short names can become ambiguous as models are added.

## See what a request cost

Non-streaming responses include an `x-everyais-cost-usd` header with the price in USD.

```python
raw = client.chat.completions.with_raw_response.create(
    model="everyais/gemini-3-8-flash",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(raw.headers.get("x-everyais-cost-usd"))
response = raw.parse()
```

## Retry the same request safely

If the connection drops before you get a response, resending can run and bill the work twice. Send an `Idempotency-Key` header and reuse it when you retry.

```python
import uuid

key = str(uuid.uuid4())  # reuse this value when you retry
response = client.chat.completions.create(
    model="everyais/gemini-3-8-flash",
    messages=[{"role": "user", "content": "Hello!"}],
    extra_headers={"Idempotency-Key": key},
)
```

- Works on Chat, Messages, Responses, image generation and image editing. Not on video submission.
- A retry matches only with the same API key, endpoint and body. Different JSON whitespace or field order is a different body.
- Keys can be up to 255 characters. Records are kept for 24 hours.
- Successful JSON responses up to 300 KiB are replayed with an `Idempotent-Replayed: true` header.
- Streams and large responses are not replayed. Resending the same key returns 409, and the request doesn't run again.
- Failed responses are not stored, so you can retry them with the same key.
- In rare cases a request may run twice.

See [errors and retries](https://platform.everyais.com/en/docs/errors-and-rate-limits#doc-409-when-an-idempotency-key-collides) for each 409 code.

## Per-endpoint limits

| Item | Limit |
|------|------|
| `n` on `/v1/chat/completions` | Only 1. 2 or more returns 400 |
| `n` on `/v1/images/generations` | 1–10 |
| `n` on `/v1/images/edits` | 1–4 |

Body size limits are in [Base URL](https://platform.everyais.com/en/docs/base-url), and per-key spend limits and scopes in [authentication](https://platform.everyais.com/en/docs/authentication). Cache token counts are covered in [prompt caching](https://platform.everyais.com/en/docs/prompt-caching).
Embeddings (`/v1/embeddings`) and audio (`/v1/audio/*`) are not available.

---