> ## Documentation Index
> Fetch the complete documentation index at: https://docs.critique-labs.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# API Studio

> Create, test, and integrate a private API from the browser

## Overview

API Studio turns a natural-language description into a private, callable research
endpoint. The manual builder is available at
[critique-labs.ai/en/design](https://critique-labs.ai/en/design).

If you want a coding agent to create and integrate the endpoint for you, select
**Set up an agent** beside the API description form. You can copy the setup
prompt, open it in ChatGPT or Claude, or download the skill package for Codex,
Claude Code, Cursor, and other clients that support agent skills.

<Note>
  The manual builder and agent workflow create the same kind of private API.
  Endpoints created by an agent appear under **Saved designs** in API Studio.
</Note>

## Creating an API

<Steps>
  <Step title="Describe Your API">
    Write a natural language description of your desired API functionality:

    * Purpose and use cases
    * Required inputs and exact output fields
    * Research behavior and preferred sources
    * How to handle missing or conflicting information
  </Step>

  <Step title="Configure Settings">
    Customize your API configuration:

    * Input request format
    * Output response format
    * Detailed API instructions
  </Step>

  <Step title="Test Your API">
    Test your API with sample requests and responses:

    * Enter an input request
    * View the output response
    * Review the returned source context
    * Revise API instructions if needed
  </Step>

  <Step title="Integrate">
    In the **Run** step, copy the endpoint URL, request schema, example payload,
    or generated JavaScript, Python, and cURL code. Keep the API key in the
    `CRITIQUE_API_KEY` environment variable.
  </Step>
</Steps>

## Build with an agent

See [Build APIs with an agent](/api-platform/agent-created-apis) for setup,
endpoint lifecycle commands, and the runtime response contract.

## Best Practices

### Writing Descriptions

* Be specific about functionality
* Name every required input
* Define the exact output shape and value types
* State freshness and source requirements
* Explain how to represent missing information

### Testing

* Validate all endpoints
* Test edge cases
* Check error responses
* Verify rate limits

## Pricing

* **\$0.50 per million tokens** (including internal processing tokens).
* Rate limits exist for the free and pro tiers. See [API rate limits](/api-reference/rate-limits) and the [pricing page](https://critique-labs.ai/en/pricing) for details.

Use the **API Reference** navigation to inspect the complete HTTP specification.