RemotionUI

MCP Server

Expose the RemotionUI registry to agents as MCP tools.

remotion-ui-mcp is a Model Context Protocol server that exposes the RemotionUI registry as agent-callable tools over stdio. It reuses the CLI's registry client, so results match remotion-ui search, view, and add exactly.

Use it when your agent should discover components on its own. For installing them, the agent still runs the CLI. The server returns the command, it does not write files.

Tools

ToolInputReturns
list-componentsregistryUrl?The full registry index
search-componentsquery?, lane?, tier?, registryUrl?Matching index entries
get-component-detailname, preset?, registryUrl?The full registry item: files, dependencies, composition, compat
get-install-commandnamenpx remotion-ui@latest add <name>

Lanes: atoms, signals, vectors, spatial, blocks, cuts, reels. Tiers: core, advanced.

Failures come back as isError results carrying the same { code, message } envelope as the CLI's --json mode. See Error codes.

Install

The server is published to npm, no checkout required.

Run
npx remotion-ui-mcp

Claude Code

Add to .mcp.json in your project root:

.mcp.json
{ "mcpServers": { "remotion-ui": { "command": "npx", "args": ["-y", "remotion-ui-mcp"] } } }

Other MCP clients

Point any stdio-capable client at the same command. Claude Desktop uses the same command / args shape in its config file.

From a checkout

Working on the server itself, or pinning to unreleased changes:

Build
git clone https://github.com/riaz37/remotion-ui cd remotion-ui && pnpm install pnpm --filter remotion-ui-mcp build

That produces packages/remotion-ui-mcp/dist/index.js. Point command: "node" at it instead.

Agent workflow with MCP

  1. search-components or list-components to find a fit.
  2. get-component-detail to read files, dependencies, and composition metadata.
  3. get-install-command, then run it with the CLI.
  4. Import from local source paths, never from the remotion-ui npm package.

The agent skill encodes the same workflow for agents that do not speak MCP. See AI Usage for the plain HTTP endpoints (/llms.txt, /ai/components.json).

On this page