FFmpeg MCP Server for Zed: Video Processing in Your AI Editor

Zed is a fast, Rust-native code editor with built-in MCP support. Connect the FFmpeg Micro MCP server and you can process video by typing a prompt in the AI assistant. No terminal switching, no curl commands to construct, no API docs to keep open in another tab.

This guide covers setup, configuration, and working examples of video processing through Zed.

Set up FFmpeg Micro MCP in Zed (2 minutes)

You need two things: a free FFmpeg Micro API key (grab one at ffmpeg-micro.com) and Zed installed.

Add the MCP server to Zed settings

Open Zed settings with Zed > Settings (or Cmd+, on macOS). Add the context_servers block:

{
  "context_servers": {
    "ffmpeg-micro": {
      "command": {
        "path": "npx",
        "args": ["-y", "@ffmpeg-micro/mcp-server"],
        "env": {
          "FFMPEG_MICRO_API_KEY": "your_api_key_here"
        }
      }
    }
  }
}

Replace your_api_key_here with your actual API key and save the file.

Verify the connection

Open Zed's AI assistant panel and start a new conversation. The FFmpeg Micro tools should appear in the available tools list. If they don't show up, restart Zed and make sure Node.js 18+ is installed (npx needs it).

What the MCP server exposes

Once connected, Zed's AI assistant can call these tools directly:

You describe what you want in plain English. The assistant figures out which tool to call.

Real examples: video processing from Zed

Open the AI assistant panel and try these prompts.

Convert MP4 to WebM:

> "Convert this video to WebM at high quality: https://storage.example.com/demo.mp4"

The assistant calls create_transcode with the right format and quality settings, returns the job ID, and can check status when you ask.

Resize for Instagram Reels:

> "Make this video vertical 1080x1920 for Instagram: https://storage.example.com/landscape.mp4"

The assistant constructs the scale filter options and sends the transcode request. No need to remember the exact -vf syntax.

Extract audio as WAV:

> "Pull just the audio from this video as WAV: https://storage.example.com/interview.mp4"

The assistant uses advanced options mode to pass -vn (no video) and sets the output format.

Generate a thumbnail:

> "Extract a frame at 5 seconds from this video: https://storage.example.com/tutorial.mp4"

The assistant uses frame extraction options to grab a single frame as an image.

Batch processing:

> "Transcode these three videos to 720p MP4: [url1, url2, url3]"

The assistant creates separate transcode jobs for each URL and tracks their status.

Why Zed for video processing

Speed. Zed is written in Rust. The editor stays responsive while MCP tools are running in the background. Process a video and keep coding without the UI freezing.

Native MCP support. MCP is built into Zed's core, not bolted on through an extension. Configuration lives in the same settings.json you already use for everything else.

Context awareness. Zed's assistant can read your open files and project context. You can reference code in your project when asking for video processing. "Transcode the sample video URL from line 42 of my config file" works because the assistant can see your code.

Minimal setup. One JSON block in settings, one API key. No extension marketplace, no plugin installs, no binary downloads. npx handles the MCP server automatically.

Upload local files for processing

If your video isn't on a public URL, upload it through the presigned URL flow:

> "I need to process a local video. Help me upload ~/Videos/raw-footage.mp4 to FFmpeg Micro."

The upload step still needs a terminal command because MCP tools can't access your local filesystem directly. But the assistant generates the exact command.

Common Pitfalls

Node.js version. The MCP server requires Node.js 18 or later. Run node --version to check. Update with your package manager or nvm if needed.

JSON syntax in settings. A missing comma or quote in settings.json breaks the config silently. Zed won't load the MCP server and won't always tell you why. Double-check your JSON if tools don't appear.

AI model selection. Zed supports multiple providers (Anthropic, OpenAI, Ollama). MCP tool use works best with Claude models. If your current provider doesn't support tool calling well, switch to Claude in Zed's AI settings.

File size limits. The free tier supports files up to 100MB. For larger files, upgrade to a paid plan. The assistant tells you if a file exceeds the limit.

FAQ

Is the FFmpeg Micro MCP server free?

The MCP server package is free and open source. FFmpeg Micro itself has a free tier with 10 minutes of video processing per month, enough for testing and small pipelines. Paid plans start at /month.

Does this work on Linux?

Yes. Zed runs on macOS and Linux with the same MCP configuration. Windows support for Zed is still in progress.

Can I use other MCP servers alongside FFmpeg Micro in Zed?

Add multiple entries to the context_servers object. Each server gets its own block. Zed's assistant can use tools from all connected servers in a single conversation.

What video formats does FFmpeg Micro support?

Input: MP4, WebM, MOV, AVI, MKV, and most formats FFmpeg can read. Output: MP4, WebM, or MOV.

How is this different from just running FFmpeg in the terminal?

You skip FFmpeg installation, codec management, and server maintenance entirely. Describe what you want in plain English, the AI builds the right FFmpeg parameters, and processing happens in the cloud. Your machine doesn't do the heavy lifting.