Overview
Lirovo speaks MCP. Connect any AI agent and video extraction becomes a tool it can call: hand it a link, get back typed JSON with a citation for every value. No SDK, no glue code. The MCP server is the same surface as the REST API, exposed over Model Context Protocol.
What MCP is for Lirovo#
The Model Context Protocol (MCP) lets an AI agent discover and call your tools directly. Lirovo's MCP server hands an agent a set of tools for turning a source video into structured data: it submits a job from a link, polls the typed result, walks the Pass A knowledge graph, and cites the exact source moment behind every extracted value. The agent runs that loop on its own. You write no SDK code and no orchestration glue.
Crucially, the MCP server is the same surface as the REST API. The tools call the same control-plane logic the REST endpoints do, so tenancy, per-tenant concurrency and daily-cost caps, validation, and delivery all behave identically. MCP is transport plus auth plus result shaping on top of the API you already know.
Endpoint and transport#
The server lives at https://api.lirovo.ai/mcpand speaks Streamable HTTP. It is built on Cloudflare's agents package (the McpAgent served at /mcp) and fronted by workers-oauth-provider. Each session is backed by a Durable Object, so the heavy work stays in the existing extraction pipeline and the MCP layer stays thin.
https://api.lirovo.ai/mcp/mcp returns 404, MCP is not yet enabled on that environment and the REST API is the surface to use in the meantime.Next steps#
Connect a client, then browse the full tool surface.