Chat Completions API

OpenAI-compatible chat completions with built-in reasoning, tool use, and streaming.

Overview

The Chat Completions API is the primary interface for interacting with MiroMind models. It follows the OpenAI Chat Completions format, so you can use the OpenAI SDK or any compatible client. The API is served by the frontier gateway.

Base URL:https://api.miromind.ai

Auth:Bearer API Key

Format:OpenAI-compatible

Models

Pass the model id below in the model parameter when creating a chat completion.

MiroThinker 1.7 Deep Research

mirothinker-1-7-deepresearch

Flagship deep research model. Stronger reasoning depth and broader tool use.

Context window256k tokens

Max completion16k tokens

MiroThinker 1.7 Deep Research Mini

mirothinker-1-7-deepresearch-mini

Faster, lower-cost variant of the flagship. Same context window with reduced reasoning depth.

Context window256k tokens

Max completion16k tokens

List via API

The same data is available programmatically — useful for clients that want to enumerate or check model capabilities at runtime.

GET/v1/models

Response

{
  "object": "list",
  "data": [
    {
      "id": "mirothinker-1-7-deepresearch",
      "object": "model",
      "created": 1700000000,
      "owned_by": "miromind",
      "context_length": 262144,
      "max_completion_tokens": 16384
    },
    {
      "id": "mirothinker-1-7-deepresearch-mini",
      "object": "model",
      "created": 1700000000,
      "owned_by": "miromind",
      "context_length": 262144,
      "max_completion_tokens": 16384
    }
  ]
}

Example

curl https://api.miromind.ai/v1/models \
  -H "Authorization: Bearer YOUR_API_KEY"

POST/v1/chat/completions

Create Chat Completion

Send a conversation to the model and receive a completion. Supports streaming (SSE) and non-streaming modes.

Headers

AuthorizationRequired

Your MiroMind API key

Bearer YOUR_API_KEY

Content-TypeRequired

Must be application/json

application/json

Parameters

modelstringRequired

Model ID to use, e.g. "mirothinker-1-7-deepresearch-mini"

messagesarrayRequired

Array of message objects with "role" (system/user/assistant) and "content" fields

streamboolean

Whether to stream the response via SSE. Defaults to true.

max_tokensinteger

Maximum number of tokens to generate in the completion

mcp_serversarray

Array of MCP server configs ({name, url, headers?, access_token?, oauth?}) for external tool access

Request

curl -X POST https://api.miromind.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mirothinker-1-7-deepresearch-mini",
    "messages": [
      {
        "role": "user",
        "content": "What are the latest trends in AI?"
      }
    ],
    "stream": true
  }'

Response

Non-Streaming

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1712345678,
  "model": "mirothinker-1-7-deepresearch-mini",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Here are the latest trends in AI..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 256,
    "total_tokens": 268
  }
}

Streaming Response

When stream: true (the default), the response is delivered as Server-Sent Events (SSE). Each line is prefixed with data: followed by a JSON chunk. The stream ends with data: [DONE]. The stream has two phases:

Phase 1: Reasoning

The model emits reasoning steps via delta.reasoning_steps. Each step has a type (e.g. thinking, web_search) and a content field with the step details.

Phase 2: Final Answer

After reasoning completes, the final answer streams via delta.content, token by token, just like standard OpenAI streaming.

SSE Stream Example

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"role":"assistant"}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"reasoning_steps":[{"type":"thinking","thought":"Let me analyze this question..."}]}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"reasoning_steps":[{"type":"web_search","web_search":{"search_keywords":["latest AI trends 2026"]}}]}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":"Here are"}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":" the latest"}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":" trends..."}}]}

data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":{"prompt_tokens":12,"completion_tokens":256,"total_tokens":268,"completion_tokens_details":{"reasoning_tokens":45},"num_search_queries":1}}

data: [DONE]

Non-Streaming Response

Set stream: false to receive the full response as a single JSON object. The response waits until the model finishes generating before returning.

{
  "id": "chatcmpl-abc123",
  "object": "chat.completion",
  "created": 1712345678,
  "model": "mirothinker-1-7-deepresearch-mini",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Here are the latest trends in AI...",
        "reasoning_steps": [
          {
            "type": "thinking",
            "thought": "Let me analyze this question..."
          },
          {
            "type": "web_search",
            "web_search": {
              "search_keywords": ["latest AI trends 2026"],
              "search_results": [{"title": "...", "url": "...", "snippet": "..."}]
            }
          }
        ]
      },
      "finish_reason": "stop"
    }
  ],
  "search_results": [{"title": "...", "url": "...", "snippet": "..."}],
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 256,
    "total_tokens": 268,
    "completion_tokens_details": {
      "reasoning_tokens": 45
    },
    "num_search_queries": 1
  }
}

Finish Reason & Usage

The final chunk (streaming) or the response (non-streaming) includes a finish_reason and a usage object.

finish_reason	Description
`stop`	The model completed normally
`error`	The workflow failed due to an internal error. An `error` object is included in the chunk.
`cancelled`	The request was cancelled (client disconnect or explicit cancel)

usage field	Description
`prompt_tokens`	Tokens in the input prompt
`completion_tokens`	Tokens generated in the completion
`total_tokens`	Sum of prompt + completion tokens
`completion_tokens_details.reasoning_tokens`	Hidden reasoning tokens used by the model to produce the answer. Counted as part of `completion_tokens` and billed as output. OpenAI-compatible nested shape.
`num_search_queries`	Number of billed `fetch_url_content` invocations during reasoning. Omitted when zero.

Reasoning Step Types

During the reasoning phase, the model can emit the following step types:

Type	Description
`thinking`	Internal reasoning and analysis
`web_search`	Searching the web for information
`fetch_url_content`	Fetching and reading content from a URL
`execute_python`	Executing Python code in a sandboxed environment
`execute_command`	Running a shell command
`tool_call`	Calling an MCP tool or external function

Capabilities & Limits

Capability	Status
Context window	256k tokens per request, shared between input and output.
External tool use (MCP)	Supported via the `mcp_servers` request parameter; currently in private beta. Contact us for access.
Custom function calling	OpenAI-style `tools` / `tool_choice` are not supported. Use `mcp_servers` for external tools.
Structured output	`response_format` / `json_schema` are not supported.
Prompt caching	`cache_control` is not supported.
Multimodal input	Image and document inputs are supported on the platform but not yet exposed via the public API. Coming soon.

OpenAI SDK Compatibility

Point any OpenAI SDK at the MiroMind base URL and use your MiroMind API key. The API is fully compatible with the OpenAI Chat Completions format.

Python (openai SDK)

from openai import OpenAI

client = OpenAI(
    base_url="https://api.miromind.ai/v1",
    api_key="YOUR_API_KEY",
)

stream = client.chat.completions.create(
    model="mirothinker-1-7-deepresearch-mini",
    messages=[
        {"role": "user", "content": "What are the latest trends in AI?"}
    ],
    stream=True,
)

for chunk in stream:
    delta = chunk.choices[0].delta
    if delta.content:
        print(delta.content, end="", flush=True)

Node.js (openai SDK)

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'https://api.miromind.ai/v1',
  apiKey: 'YOUR_API_KEY',
});

const stream = await client.chat.completions.create({
  model: 'mirothinker-1-7-deepresearch-mini',
  messages: [
    { role: 'user', content: 'What are the latest trends in AI?' }
  ],
  stream: true,
});

for await (const chunk of stream) {
  const content = chunk.choices[0]?.delta?.content;
  if (content) process.stdout.write(content);
}

Error Responses

The API returns standard HTTP error codes with a JSON error body.

Status	Code	Description
400	bad_request	Invalid request body or missing required fields
401	unauthorized	Missing or invalid API key
402	insufficient_balance	Account balance is too low to process the request
429	rate_limited	Too many requests. Retry after the Retry-After header value.
503	service_unavailable	The service is temporarily overloaded or down for maintenance

Error Response Format

{
  "error": {
    "code": "unauthorized",
    "message": "Invalid API key provided.",
    "type": "authentication_error"
  }
}

Next Steps

For long-running tasks that should survive disconnects, use the Responses API.

Responses API Streaming Guide

Chat Completions API

MiroThinker 1.7 Deep Research

MiroThinker 1.7 Deep Research Mini

List via API

Response

Example

Headers

Parameters

Request

Response

Phase 1: Reasoning

Phase 2: Final Answer

SSE Stream Example

Python (openai SDK)

Node.js (openai SDK)

Error Response Format

On this page