PREMIERE PRO MCP DOCUMENTATION

AI editing tools for Adobe Premiere Pro

Premiere Pro MCP is an open-source, local-first Model Context Protocol server that connects AI assistants to Adobe Premiere Pro through 269 structured editing, inspection, automation, and export tools.

How to install Premiere Pro MCP

  1. Install Node.js 18 or newer on Windows or macOS.
  2. Run npm install -g premiere-pro-mcp.
  3. Run premiere-pro-mcp --install-cep.
  4. Configure Claude Desktop, Cursor, Windsurf, or another MCP client to launch premiere-pro-mcp.
  5. Open Premiere Pro and call ping to verify the live bridge.

What can an AI assistant do in Premiere Pro?

Timeline editing

Insert, overwrite, move, trim, split, ripple-delete, target tracks, and inspect sequence structure.

Effects and color

Apply effects, control Lumetri properties, add keyframes, use LUTs, and verify resulting values.

Media and projects

Import footage, organize bins, create sequences, inspect metadata, manage proxies, and save projects.

Audio and captions

Adjust verified audio levels, automate keyframes, mute tracks, create captions, and inspect audio state.

Export

Discover Adobe Media Encoder presets, export sequences and project items, and verify frame output on disk.

Inspection and workflows

Use 269 tools, three resources, four prompts, edit-plan previews, capability profiles, and audit events.

Windows, macOS, CEP, and UXP compatibility

The packaged CEP bridge targets Premiere Pro 2020 through 2026 on Windows and macOS. The UXP bridge is a preview for Premiere Pro 25.6 and newer. Compatibility is not the same as a live connection: always call get_capabilities and ping on the editor's machine before running a workflow.

Local-first architecture and security

The recommended stdio setup keeps Premiere Pro, the MCP server, and project media on the local computer. Capability profiles separate inspection, editing, export, filesystem access, and unsafe scripting. Raw ExtendScript is disabled by default and requires explicit operator authority.

Canonical resources