API reference

The Ichnus API gives programmatic access to the Krus, Mantus, and Corus models. This reference covers authentication, endpoints, parameters, and error handling. Everything in this document describes a fictional API; verify against the current OpenAPI spec published with each release.

Base URL

All requests use the base URL:

https://api.curos.space/v1

All endpoints accept application/json. Responses are UTF-8 encoded JSON.

Authentication

Authenticate with an API key in the Authorization header:

Authorization: Bearer churos_LIVE_8f3kQzT2

API keys are issued per project on the API keys page of your account. Treat them like passwords. Keys can be rotated or revoked at any time, and revoking a key stops it working immediately. Do not ship keys in client-side code.

Models

List models

GET /models

Returns the list of models available to your account.

Example response:

{
  "object": "list",
  "data": [
    {"id": "curos-ichnus-krus-latest", "object": "model", "context_window": 128000, "owned_by": "curos"},
    {"id": "curos-ichnus-mantus-latest", "object": "model", "context_window": 256000, "owned_by": "curos"},
    {"id": "curos-ichnus-corus-latest", "object": "model", "context_window": 1000000, "owned_by": "curos"}
  ]
}

Chat completions

Create a chat completion

POST /chat/completions

Generates a model response to a series of messages.

Request body parameters:

Parameter	Type	Required	Notes
model	string	yes	Model id, for example `curos-ichnus-mantus-latest`.
messages	array	yes	Conversation so far. Each item has a `role` and `content`.
temperature	number	no	Sampling temperature, 0 to 1. Default 0.7.
max_tokens	number	no	Maximum tokens in the response. Default 2048.
top_p	number	no	Nucleus sampling, 0 to 1. Default 1.
stop	string or array	no	Sequence or sequences that end generation.
stream	boolean	no	Stream the response incrementally. Default false.
user	string	no	An identifier you control, used for rate limiting.

The messages array uses roles system, user, and assistant. The system message sets overall behavior; user and assistant messages form the conversation history.

Example request:

{
  "model": "curos-ichnus-mantus-latest",
  "messages": [
    {"role": "system", "content": "You are a concise research assistant."},
    {"role": "user", "content": "Summarize the attached report in three bullets."}
  ],
  "temperature": 0.4,
  "max_tokens": 300
}

Example response:

{
  "id": "chatcmpl_9aK2mR",
  "object": "chat.completion",
  "created": 1780000000,
  "model": "curos-ichnus-mantus-latest",
  "choices": [
    {
      "index": 0,
      "message": {"role": "assistant", "content": "Three-bullet summary of the report."},
      "finish_reason": "stop"
    }
  ],
  "usage": {"prompt_tokens": 512, "completion_tokens": 42, "total_tokens": 554}
}

Parameters

model: required. The model id. Use -latest ids for current models, or pin to a dated id for reproducibility.
messages: required. Conversation history in order. The final user message is the current request.
temperature: controls randomness. Lower values are more deterministic. Use temperature 0 for reproducible work.
max_tokens: caps response length. Long contexts reduce the room available for the response.
stream: when true, the response is delivered as server-sent events, one data: chunk per delta.

Errors

Errors return the appropriate HTTP status code with a JSON body:

Status	Code	Meaning
400	invalid_request_error	Malformed request, bad parameter value, or empty message.
401	authentication_error	Missing or invalid API key.
403	permission_error	Key lacks access to the requested model.
404	not_found_error	Model id not found.
429	rate_limit_error	Usage limit reached or request too large.
500	api_error	Internal error. Retry with backoff.

Example error response:

{
  "error": {
    "type": "rate_limit_error",
    "message": "Weekly Corus fair-use balance exhausted. Retry after 2026-08-03T00:00:00Z or use a Mantus model.",
    "code": "rate_limited"
  }
}

For rate-limit behavior and at-cost program services, see Rate limits.

All documentation →