High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.
MCP server for justfile tasks
just-mcp connects a justfile to an agent through MCP so the agent can list, understand, and run project tasks. It parses recipes, watches for file changes, and publishes them as tools the agent can call.
Builders who want their agent to reuse the same justfile tasks across project sessions.
You can ask your agent to run the right project task instead of retyping shell commands or explaining your workflow again.
What it does
Justfile discovery and hot reload
Finds a project justfile, watches it for changes, and refreshes the tool list automatically.
AST-based recipe parsing
Uses Tree-sitter parsing first, with CLI and regex fallbacks, to read task names, parameters, dependencies, and comments.
Dynamic MCP tools
Turns each just recipe into a discoverable MCP tool the agent can call.
Admin tools
Includes `admin_sync` to refresh the registry and `admin_create_task` to help create a new task with backup.
Security checks
Validates inputs, limits resources, and constrains directories to reduce command injection risk.
Optional vector search
Adds offline semantic search for finding similar automation patterns across projects.
How to get it
- 11. Install just-mcp
# Using cargo-binstall (fastest - downloads pre-built binary) cargo binstall --git https://github.com/toolprint/just-mcp just-mcp # Or download latest release curl -L https://github.com/toolprint/just-mcp/releases/latest/download/just-mcp-$(uname -m)-$(uname -s).tar.gz | tar xz sudo mv just-mcp /usr/local/bin/
- 23. Use the Slash Command
/just:do-it build the project
- 3Run
claude mcp add -s user -t stdio just -- just-mcp
- 4Run
git clone https://github.com/toolprint/just-mcp.git cd just-mcp just quickstart # Complete setup + install
- 5Or manually
cargo install --path . --features all # All features including vector search
- 6Run
just quickstart # Complete setup for new developers just dev-setup # Comprehensive development environment
README
just-mcp
Transform justfiles into AI-accessible automation tools through the Model Context Protocol
just-mcp bridges justfiles and Coding Agents by exposing justfile recipes as dynamically discoverable MCP tools.
This enables AI assistants to understand, explore, and execute a project's common development workflows similar to how a human would.
Quick Start
1. Install just-mcp
# Using cargo-binstall (fastest - downloads pre-built binary)
cargo binstall --git https://github.com/toolprint/just-mcp just-mcp
# Or download latest release
curl -L https://github.com/toolprint/just-mcp/releases/latest/download/just-mcp-$(uname -m)-$(uname -s).tar.gz | tar xz
sudo mv just-mcp /usr/local/bin/
2. Configure your Agent (see MCP Client Setup below)
3. Use the Slash Command:
/just:do-it build the project
It will find the appropriate justfile task to run as an MCP Tool.
That's it. Your AI can now use your justfile tasks.
MCP Client Setup
Claude Code
claude mcp add -s user -t stdio just -- just-mcp
Or manually add to ~/.claude.json:
{
"mcpServers": {
"just": {
"type": "stdio",
"command": "just-mcp"
}
}
}
Verification: Start a new session and check /mcp for 'just' and view details to see how many tools loaded.
Claude Desktop
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"just": {
"command": "just-mcp",
"args": ["--watch-dir", "/Users/username/workspace/your-project-dir"],
}
}
}
Verification: Restart Claude Desktop, check the settings button for custom connection tools.
Cline (VS Code Extension)
- Click Cline icon → MCP Servers → Configure MCP Servers
- Add configuration:
{
"mcpServers": {
"just": {
"disabled": false,
"timeout": 60,
"type": "stdio",
"command": "just-mcp",
"args": [
"-w",
"/Users/username/your-project-dir"
]
}
}
Verification: Look for MCP server status in Cline's interface.
NOTE: Since cline runs outside of your current project working directory you MUST specify the watch-dir or use the admin tools at runtime to ask Cline to switch the watch-dir for you.
Key Features
🔍 Smart Justfile Discovery
- Real-time monitoring with hot reloading
- Defaults to use the current project root directory to look for your justfile
- Dynamic tool generation from your tasks
📝 Advanced Parsing
- AST-based parser using Tree-sitter for complete syntax support
- Parameter documentation from comments:
# {{param}}: description - Three-tier fallback: AST → CLI → Regex for maximum compatibility
🛡️ Security First
- Input validation prevents command injection
- Configurable timeouts and resource limits
- Directory whitelisting and parameter sanitization
⚙️ Admin Tools
admin_sync: Refresh tool registryadmin_create_task: AI-assisted task creation with backup
🔍 Vector Search (Optional)
- Offline semantic search with local embeddings (no API keys)
- Natural language queries: "deploy to production"
- Cross-project discovery of similar automation patterns
- See Vector Search docs
Example Workflow
Given this justfile:
# Deploy the application
deploy env="prod":
echo "Deploying to {{env}}"
./deploy.sh {{env}}
# Run tests with coverage
test coverage="false":
cargo test {{if coverage == "true" { "--coverage" } else { "" }}}
Your AI assistant can:
- Discover: "What tasks are available?"
- Execute: "Deploy to staging" → runs
deploy env="staging" - Understand: Sees parameters, descriptions, and dependencies
Installation Options
Pre-built Binaries (Recommended)
Download from GitHub Releases:
# Linux x86_64
curl -L https://github.com/toolprint/just-mcp/releases/latest/download/just-mcp-x86_64-unknown-linux-gnu.tar.gz | tar xz
# Linux ARM64
curl -L https://github.com/toolprint/just-mcp/releases/latest/download/just-mcp-aarch64-unknown-linux-gnu.tar.gz | tar xz
# macOS (Universal - works on Intel and Apple Silicon)
curl -L https://github.com/toolprint/just-mcp/releases/latest/download/just-mcp-universal2-apple-darwin.tar.gz | tar xz
From Source
git clone https://github.com/toolprint/just-mcp.git
cd just-mcp
just quickstart # Complete setup + install
Or manually:
cargo install --path . --features all # All features including vector search
Development
Setup
just quickstart # Complete setup for new developers
just dev-setup # Comprehensive development environment
Workspace Structure
./- Main MCP serverdev-tools/- Performance analysis utilities
Common Commands
just build # Build main server
just build-dev-tools # Build development utilities
just test # Run tests
just check # Format, lint, test (pre-commit)
Multi-Project Example
Monitor multiple projects with custom names:
just-mcp \
--watch-dir ~/projects/api:backend \
--watch-dir ~/projects/web:frontend \
--watch-dir ~/infrastructure:infra
Tools will be available as:
just_deploy@backendjust_build@frontendjust_apply@infra
Documentation
- Configuration Guide - Detailed setup for all MCP clients
- Vector Search - Semantic search with natural language
- Troubleshooting - Common issues and solutions
- Contributing - Development workflow and guidelines
Architecture
Justfile Watcher → AST Parser → Tool Registry → MCP Server → AI Assistant
↓ ↓ ↓ ↓
File Changes → Dynamic Updates → Real-time → Task Execution
Key Design Principles:
- Async-first: Tokio-based for concurrent operations
- Security-focused: All inputs validated, resources limited
- Hot-reload: File changes trigger automatic tool updates
- Zero-config: Works with existing justfiles
Contributing
We welcome contributions! See CONTRIBUTING.md for guidelines.
Quick steps:
just quickstart- Set up development environment- Make your changes and add tests
just check- Ensure code quality- Submit a Pull Request
License
MIT License - see LICENSE for details.
Acknowledgments
- just - The command runner that inspired this project
- Model Context Protocol - Enabling AI-tool communication
- Anthropic - For developing the MCP standard
Need help? Check troubleshooting or open an issue on GitHub.
Files in the repo
- .claude
- .dagger
- .github
- .serena
- .taskmaster
- assets
- benches
- demo
- dev-tools
- docs
- just
- src
- tests
- .gitignore
- .markdownlint.json
- .mcp.json
- .prettierignore
- .prettierrc
- Brewfile
- Cargo.toml
- CHANGELOG.md
- CLAUDE.md
- CODE_OF_CONDUCT.md
- CONTRIBUTING.md
- dagger.json
- justfile
- LICENSE
- README.md
- rust-toolchain.toml
- test_config_resource.sh
- test_justfile
- test-utilities.just
Discussion (0)
Ask about usage, or say what you built with itSign in to join the discussion.
No comments yet. Be the first to say what this is good for.
More connectors

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code
Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zero external databases or dependencies. Built in pure Go.
Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codebase so AI coding tools read only what matters, with benchmarked context reductions on reviews and large-repo workflows.
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is the proving ground. MCP server + CLI.
20 MB lightweight cross-platform database client for 90+ databases, including MySQL, PostgreSQL, SQLite, Redis, MongoDB, DuckDB, SQL Server, and Dameng. Built-in AI, MCP Server, CLI, desktop and Docker. | 轻量级跨平台数据库管理工具,支持 MySQL、PostgreSQL、SQLite、Redis、MongoDB、达梦等 90+ 数据库,提供桌面端、Docker、CLI、内置 AI 助手和 MCP Server。