Concat

What Concat does

Concat MCP server gives AI agents a way to work with the Concat desktop video editor. Concat runs on macOS, Windows, Linux, and Android, with iOS and iPadOS builds available for sideloading. Its editor supports multi-track timelines, cutting and trimming, overlays, titles, effects, transitions, keyframes, captions, background removal, voice processing, text-to-speech, and export up to 4K at 60 frames per second.

The MCP interface is intended for agent-driven editing. It complements Concat’s JSON-RPC API, which is available for scripts. A build from source also adds a command-line tool and gRPC support.

How it works

Start Concat with the --mcp option to run its built-in MCP server. An MCP-compatible agent can then send editing requests to the local application. The workflow keeps media on the computer rather than uploading footage to a remote service.

Concat also downloads its captioning, voice, and cutout models from Settings. After those models are installed, they can operate offline. This supports workflows that need local processing for source footage and generated media.

The editor can work with several cuts in one project and multiple video, still-image, and title tracks. Its editing functions include split, trim, ripple, merge, freeze-frame, speed changes, picture-in-picture, blending, cropping, and keyframed properties such as position, scale, rotation, opacity, volume, and effect parameters.

Setup and configuration

Install a Concat build for the target operating system from the project’s website or release packages, then launch it with:

concat --mcp

The MCP setup is intended for AI agents that can connect to a locally running MCP process. Models for captions, voices, and background removal are obtained through Concat’s Settings screen and remain available for offline use after downloading.

For development builds, the project requires Rust 1.93 or newer, FFmpeg 7 or newer development libraries, CMake, and a C++ compiler. The source tree can be run with cargo run -p concat from the src directory.

Tools and capabilities

• Exposes local Concat video editing to AI agents through MCP.
• Provides a JSON-RPC API for script-based editing.
• Supports multi-track and multi-timeline projects.
• Handles captions, text-to-speech, voice cloning, and background removal locally.
• Applies effects, filters, transitions, titles, and keyframe animation.
• Exports H.264, HEVC, or AV1 video, including 4K 60 fps and 10-bit color output.

Limitations and notes

Concat is in beta, so editing workflows may still have unresolved issues. AI models must be downloaded before their related local features can be used offline. Windows ARM builds are paused, and iOS or iPadOS use requires sideloading. Source builds depend on the required Rust, FFmpeg, CMake, and C++ toolchain versions.