Livepeer Agent

Overview

Livepeer Agent is the network's own agent surface: one MCP server that gives any agent harness the ability to generate, edit, and finish moving image and audio. Point your agent's connector settings at https://livepeer.org/api/mcp and the tools appear alongside whatever else it already has.

It is aimed at people who are already working in an agent — in an editor, a chat harness, or their own product — rather than at a separate web app to switch into. The network answers on the other side with models, compute, and a growing library of production skills, so the agent gains capabilities instead of the user gaining another tab.

What you can build

• Generation across media — image, video, audio, and 3D from a single connector, without wiring up a model provider per modality
• Editing and finishing — background removal, colour grading, overlays, subtitle burn-in, concatenation, and muxing, exposed as tools the agent can call in sequence
• Playbooks — production-ready multi-step workflows that run in the Agent Console rather than being reassembled from scratch each time
• Agent capabilities in your own product — an API key adds the same image and video workflows to an app you already ship

Developer surface

• How it runs — a hosted MCP server; add the URL to any agent harness that speaks the protocol, no SDK to install
• What ships with it — hundreds of capabilities spanning generation, editing, rendering, CAD, and speech, all callable as MCP tools
• How to extend it — API keys for embedding the workflows in your own product, and a playbook library for packaging repeatable multi-step jobs

Powered by Livepeer

Every job routes to the open network of independent compute providers rather than a single vendor's cluster. That is what lets one connector cover this much surface area: capabilities are contributed and served by orchestrators across the network, so the tool list grows without the agent needing a new integration each time.