Transform any Git repository into intelligent MCP tools automatically.
MCPify analyzes code repositories and generates Model Context Protocol (MCP) servers that expose repository functionality as callable tools. Simply provide a GitHub URL, describe what APIs you need, and MCPify will create a ready-to-use MCP server.
# Install MCPify
pip install mcpify
# Transform a repository into MCP tools
mcpify https://github.com/sqlite/sqlite --api "database creation and query operations"
# The generated MCP server is now ready to use with any MCP clientInput: GitHub/Git repository URL + API requirements description Output: Running MCP server with callable tools
- Repository Analysis: Clone and analyze the target repository
- Project Understanding: Extract dependencies, entry points, documentation
- API Design: Use AI to understand your requirements and map them to code
- MCP Generation: Create tools with proper schemas and implementations
- Server Launch: Start the MCP server ready for integration
Git Repository -> Static Analysis -> Semantic Layer -> Code Generator -> MCP Server Runtime -> UX Layer
Components:
- Ingestion: Repository cloning and metadata extraction
- Analysis: AST parsing, symbol graphs, documentation extraction
- Semantic: LLM-powered code understanding and API design
- Generation: MCP server scaffolding and tool creation
- Runtime: Tool execution, subprocess management, security sandbox
- Interface: CLI, web UI, VSCode extension
- Multi-language support: Python, JavaScript/TypeScript, Go, Rust
- Intelligent analysis: AST parsing + LLM-powered code understanding
- Automatic tool generation: Function signatures to MCP tool schemas
- MCP SDK integration: Official MCP SDK for guaranteed protocol compliance
- Production ready: Enterprise-grade servers with comprehensive error handling
- Safe execution: Sandboxed subprocess management with timeouts
- Hot reload: Dynamic tool updates without server restart
- Interactive UI: Web interface and VSCode extension
- Advanced security: Fine-grained permissions and resource limits
- Caching: Incremental analysis and vector indexing
- Multi-repo: Composite tools across multiple repositories
- Monitoring: Execution metrics and error tracking
- Python 3.8+
- Git
- Language-specific tools for target repositories (pip, npm, go, cargo, etc.)
pip install mcpifygit clone https://github.com/yourorg/mcpify
cd mcpify
pip install -e ".[dev]"# Transform a Python repository
mcpify analyze https://github.com/sqlite/sqlite --api "database operations"
# Transform with specific language hint
mcpify analyze https://github.com/expressjs/express --api "web server utilities"
# Custom output directory
mcpify analyze https://github.com/redis/redis --output ./my-redis-tools --api "key-value operations"# Use Anthropic Claude instead of OpenAI
mcpify analyze ./repo --api "data processing" --llm anthropic
# Generate only, don't start server
mcpify analyze ./repo --api "file operations" --no-start --verbose
# Custom model and output
mcpify analyze ./repo --api "utilities" --model gpt-4 --output ./my-toolsMCPify generates production-ready MCP servers using the official MCP SDK, providing:
- Full MCP protocol compliance: Guaranteed compatibility with all MCP clients
- Official SDK maintenance: Automatic updates and new feature support
- Rich error handling: Comprehensive error reporting and type safety
- Future-proof: Support for upcoming MCP features (resources, prompts)
All generated servers use the latest MCP SDK for maximum reliability and compatibility.
# api-config.yaml
apis:
- name: create_database
description: "Create a SQLite database and return the path"
impl:
type: "python_function"
module: "src/db/utils.py"
function: "create_db"
input_schema:
type: object
properties:
path: {type: string}
required: [path]
- name: execute_query
description: "Execute SQL query on database"
impl:
type: "python_function"
module: "src/db/utils.py"
function: "execute_sql"
input_schema:
type: object
properties:
db_path: {type: string}
query: {type: string}
required: [db_path, query]mcpify/
├── core/
│ ├── ingestion/ # Repository cloning and metadata
│ ├── analysis/ # Static analysis and AST parsing
│ ├── semantic/ # LLM-powered understanding
│ └── generation/ # MCP server generation
├── runners/ # Language-specific execution
├── security/ # Sandboxing and permissions
├── ui/ # CLI and web interfaces
└── tests/ # Test suites
# Run all tests
pytest
# Run with coverage
pytest --cov=mcpify
# Run integration tests
pytest tests/integration/- Python repository support
- Basic AST parsing with tree-sitter
- Simple LLM-based API generation
- MVP MCP server generation
- JavaScript/TypeScript support
- Go and Rust support
- Robust build/run detection
- Universal subprocess runner
- DSL for API configuration
- Static analysis + runtime validation
- Sandboxing (nsjail, Docker, WASM)
- Permission management
- Web UI and VSCode extension
- Multi-repository caching
- Vector indexing for large codebases
- Monitoring and analytics
MCPify runs code analysis and execution with several security measures:
- Sandboxed Execution: All code runs in isolated environments
- Resource Limits: CPU, memory, and time constraints
- Permission Controls: Explicit user consent for potentially risky operations
- Code Review: Generated tools include source code references for inspection
- Python: Virtual environments (venv/uv)
- Node.js: npm/pnpm isolation
- Docker: Container-based execution
- System: nsjail/firejail integration
We welcome contributions! Please see CONTRIBUTING.md for guidelines.
git clone https://github.com/yourorg/mcpify
cd mcpify
pip install -e ".[dev]"
pre-commit install
pytestMIT License - see LICENSE file for details.
- Model Context Protocol for the MCP specification
- tree-sitter for language parsing
- GitHub Linguist for language detection
- Issue Tracker: https://github.com/camel-ai/mcpify/issues
- Discussions: https://github.com/camel-ai/mcpify/discussions
- Email: [email protected]
