This MCP server enables AI assistants to interact with Jujutsu (jj) repositories through a dedicated MCP interface. It exposes a set of tools that map to common version-control-like actions in jj, such as status, log, diff, describe, and new, along with workflow helpers like bookmark_create, push, sync, and rebase. The workflow emphasizes that jj is not Git: there is no staging area, @ represents the current working commit, and describe only labels changes without moving the commit pointer. The server allows an AI to query repository state, examine history, stage and label work, create new empty commits, and manage bookmarks for collaboration. Use the provided tools to inspect and modify the repository in a structured sequence, guided by the jj workflow example in the README.

Prerequisites:

• Jujutsu (jj) installed and available in PATH
• Rust toolchain installed (for building from source)

Install steps:

1. Build from source (recommended if you have the repository): cargo build --path .

   This produces the jj-mcp-server binary, typically at target/debug/jj-mcp-server or target/release/jj-mcp-server

2. Install the binary system-wide (optional): cargo install --path .
3. Run the MCP server (example): ./target/release/jj-mcp-server

   or simply run the installed binary if in PATH

Configuration is provided in the README under the Configuration section. Create or edit the file at ~/.claude/mcp.json to register the server under the name you choose (e.g., "jj").

Environment and usage tips:

• Ensure jj is installed and available in PATH before starting the MCP server.
• The MCP server binary is named jj-mcp-server; make sure it is executable.
• The configuration uses the server name as the key (e.g., "jj"). You can modify the command if you build the binary with a different name or path.
• When using the workflow, remember that there is no staging area in jj; changes go directly into the current commit (@).
• The tools map to common actions: status, log, diff, describe, new, bookmark_create, push, sync, rebase. Use them in appropriate order to manage work.
• If you encounter PATH issues, run with an absolute path to the binary or ensure the directory containing jj-mcp-server is in PATH.
• For multi-repo setups, you may configure multiple mcpServers entries with different commands—each server runs independently.