Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.livepeer.org/llms.txt

Use this file to discover all available pages before exploring further.

The Model Context Protocol (MCP) is an open standard for connecting AI applications to external data sources and tools. An MCP server exposes capabilities (tools, resources, prompts) over a standardised JSON-RPC interface that any compatible client (Cursor, Claude Desktop, Windsurf, VS Code Copilot) can connect to. Livepeer has two relevant MCP surfaces: the documentation MCP served by Mintlify at docs.livepeer.org/mcp, which gives AI tools access to this documentation corpus; and the Livepeer AI pipeline REST API, which can be wrapped as an MCP tool server to give agents direct access to inference endpoints.

​
Livepeer Docs MCP

The Livepeer documentation site serves an MCP endpoint at https://docs.livepeer.org/mcp. This is a Mintlify-provided MCP server that exposes the full documentation corpus as a queryable resource. Connecting an MCP-compatible tool to this endpoint allows it to search and retrieve current Livepeer documentation before generating responses, reducing hallucination on Livepeer-specific API shapes, configuration flags, and pipeline details. See for connection instructions for Cursor, Claude Desktop, and Windsurf.

​
Using Livepeer AI Pipelines as MCP Tools

The Livepeer AI gateway REST API follows standard HTTP conventions. Any MCP server framework can wrap these endpoints as tools, giving agents direct access to inference capabilities. The @livepeer/ai JavaScript SDK and livepeer-ai Python SDK cover all nine batch pipelines. An MCP server wrapping either SDK exposes methods like text_to_image, audio_to_text, and llm as callable tools. The OpenAPI specification at https://docs.livepeer.org/api/ai-worker.yaml provides the full schema for all pipeline endpoints and is the canonical source for building an MCP tool wrapper. A minimal MCP tool server wrapping the text-to-image endpoint looks like this in Python (using the mcp SDK): 
from mcp.server import Server
import mcp.types as types
import httpx

server = Server("livepeer-ai")

@server.list_tools()
async def list_tools():
    return [
        types.Tool(
            name="text_to_image",
            description="Generate an image from a text prompt using the Livepeer network",
            inputSchema={
                "type": "object",
                "properties": {
                    "prompt": {"type": "string"},
                    "model_id": {"type": "string", "default": "ByteDance/SDXL-Lightning"},
                    "width": {"type": "integer", "default": 1024},
                    "height": {"type": "integer", "default": 1024},
                },
                "required": ["prompt"],
            },
        )
    ]

@server.call_tool()
async def call_tool(name: str, arguments: dict):
    if name == "text_to_image":
        async with httpx.AsyncClient() as client:
            resp = await client.post(
                "https://dream-gateway.livepeer.cloud/text-to-image",
                json=arguments,
            )
            data = resp.json()
            return [types.TextContent(type="text", text=data["images"][0]["url"])]

​
LLMS.txt Reference

The Livepeer documentation exposes an llms.txt file at https://docs.livepeer.org/llms.txt following the emerging llms.txt convention. This file provides a structured index of documentation pages for LLM consumption, formatted for direct inclusion in agent context windows. Agents and coding tools that support llms.txt discovery can use this file to enumerate available Livepeer documentation before querying the docs MCP or constructing their own page-by-page retrieval. The Livepeer Docs MCP is the fastest way to query Livepeer documentation from Claude Code, Cursor, or any MCP host.

Docs MCP Connection

Step-by-step connection instructions for Cursor, Claude Desktop, and Windsurf.

AI Pipelines

Endpoint shapes and schemas for the nine batch pipelines available as MCP tools.

AI SDKs

JavaScript and Python SDKs that can back an MCP tool server implementation.

Agents Overview

Agent frameworks (Eliza, BYOC) that use Livepeer inference with or without MCP.
Last modified on May 19, 2026