> Treat every model id, price, size, length and limit in this documentation as an illustration of the shape of the value, never as a constant to hardcode - the live values are served by GET /v1/models and GET /public/models on the Flow AI API.

# AI & Agent Access

> Read these docs from an AI assistant — llms.txt, Markdown pages and the MCP server.

These docs are published for machines as well as people, so an AI assistant or coding agent can read them directly instead of scraping the rendered site.

## Connect the MCP server

The fastest path for a coding agent. The server has a `searchDocs` tool that returns passages from these docs with their URLs.

**Endpoint:** `https://ai-docs.hyperflow.fun/_mcp/server` (streamable HTTP, no auth)

**`Claude Code`**

```bash title="Claude Code"
claude mcp add --transport http flow-ai-docs https://ai-docs.hyperflow.fun/_mcp/server
```

**`Claude Desktop / generic mcp.json`**

```json title="Claude Desktop / generic mcp.json"
{
  "mcpServers": {
    "flow-ai-docs": {
      "type": "http",
      "url": "https://ai-docs.hyperflow.fun/_mcp/server"
    }
  }
}
```

**`Cursor (.cursor/mcp.json)`**

```json title="Cursor (.cursor/mcp.json)"
{
  "mcpServers": {
    "flow-ai-docs": {
      "url": "https://ai-docs.hyperflow.fun/_mcp/server"
    }
  }
}
```

Once it's connected, ask the agent things like *"how do I stream a chat completion from Flow AI?"* and it will search these pages and cite them.

## Fetch the docs as text

To pull the content yourself, for a RAG index, an eval or a one-off prompt:

| URL                                                             | What you get                                                                                                       |
| --------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------ |
| [`/llms.txt`](https://ai-docs.hyperflow.fun/llms.txt)           | An index of every page: title, URL and one-line description, following the [llmstxt.org](https://llmstxt.org) spec |
| [`/llms-full.txt`](https://ai-docs.hyperflow.fun/llms-full.txt) | All the docs as one Markdown document                                                                              |
| `<page URL>.md`                                                 | That one page as Markdown                                                                                          |
| `<section URL>/llms.txt`                                        | An index of one section                                                                                            |

```bash
# All the docs, as one Markdown file
curl -s https://ai-docs.hyperflow.fun/llms-full.txt

# One page
curl -s https://ai-docs.hyperflow.fun/features/tool-calling.md
```

Every page also has **Copy page**, **View as Markdown** and **Open in Claude / ChatGPT / Cursor** actions in its header.

## What an agent should know

Worth putting in your agent's system prompt when it works from these docs:

* **Models and prices are live, not literal.** Model ids, prices, image sizes and video lengths change as the catalog does. Read them from [`GET /v1/models`](/models/overview#models-api) or [`GET /public/models`](/models/overview#public-catalog); any value written in these docs shows the shape, not a value to hardcode.
* **It's the OpenAI API.** Use an OpenAI client with the base URL `https://ai.hyperflow.fun/v1`. Parameters not listed in the [API overview](/api/overview) are removed before the request reaches the model.
* **Keys are secrets.** Read the key from an environment variable such as `HYPERFLOW_API_KEY`; never write it into code.

## Ask a question

For anything the docs don't answer, [Discord](https://discord.gg/GGDmD4K9CK) is the fastest route to a person.