kodit

Kodit is an MCP server that indexes and exposes code snippets from your codebases to your AI coding assistant. Once running, it builds a searchable index of local and remote repositories, supports keyword and semantic search, and provides rich context parameters and metadata to improve the accuracy of code examples returned by your assistant. It integrates with popular AI coding assistants and can be deployed alongside other enterprise tooling. To use Kodit, start the MCP server and connect your coding assistant to the Kodit MCP endpoint. You can then issue queries like “show me Python examples of REST clients” or filter results by language, repository, author, or date range. The server also supports hybrid search (BM25 with semantic) and advanced snippet extraction with context-aware dependencies to reduce hallucinations in generated code.

Prerequisites:

• Node.js (14.x or newer) and npm
• Git
• Optional: Docker if you prefer containerized deployment

1. Clone the repository

• git clone https://github.com/helixml/kodit.git
• cd kodit

2. Install dependencies

• npm install

3. Configure environment

• Copy example env and adjust values
  • cp .env.example .env
  • Edit .env to set OPENAI_API_KEY, DATABASE_URL (if using an external DB), PORT, INDEX_PATH, etc.

4. Run the server

• npm run build (if a build step exists in the project)
• npm start

5. (Optional) Run with Docker

• Create a docker-compose.yaml or run a container image if provided by the project maintainers.
• Example (adjust image name and environment as needed): docker run -p 3000:3000 -e OPENAI_API_KEY=your_key -e INDEX_PATH=/code/index kodit:latest

6. Verify

• Access the MCP endpoint at http://localhost:3000/mcp (or the port you configured) and try a sample query via your coding assistant integration.

Tips and common considerations:

• Environment variables: Ensure OPENAI_API_KEY (or other model provider keys) are set if you rely on external models.
• INDEX_PATH and DATABASE_URL: Use a persistent store for the index to avoid data loss between restarts. For heavy usage, consider a dedicated database or vector store as recommended by the docs.
• Privacy and security: Respect repository access controls when indexing private repos (use PATs or appropriate credentials).
• Integration: Kodit is designed to work with multiple MCP clients; verify the client-specific integration steps in the hosted docs for Cursor, Cline, Claude Code, etc.
• Updates: Regularly reindex to keep snippets up-to-date with source changes.
• Performance: For large codebases, consider incremental or scheduled reindexing to manage resource usage.
• Logging: Increase log level to debug while troubleshooting; reduce after issues are resolved to improve performance.