mcp-clj

What is mcp-clj?

mcp-clj is a Clojure implementation of the Model Context Protocol (MCP) with minimal dependencies and self-contained Clojure REPL integration. It provides both MCP server and client implementations, supporting multiple transports (SSE, stdio, in-memory) and built-in tools like clj-eval and ls — ideal for Clojure developers who want to expose REPL functionality to MCP clients like Claude Desktop or build Clojure applications that consume MCP services.

How to use mcp-clj?

Add the git dependency to your deps.edn and start the server via clj -M:stdio-server (stdio) or clj -M:sse-server (HTTP). Programmatically, call mcp-clj.mcp-server.core/create-server with a transport configuration (e.g., {:transport {:type :stdio}}). For Claude Desktop or Claude Code, configure the stdio transport in the respective config file or CLI. The client API connects to other MCP servers from Clojure using create-client, list-tools, and call-tool.

Key features of mcp-clj?

• Self-contained Clojure REPL evaluation (no nREPL required)
• Low dependencies: only org.clojure/data.json
• Multiple transports: SSE (HTTP), stdio, and in-memory
• Built-in tools: clj-eval and ls with gitignore support
• MCP client for connecting to other MCP servers
• Dynamic tool addition via add-tool!

Use cases of mcp-clj?

• Expose Clojure REPL to Claude Desktop for interactive code evaluation
• Build Clojure applications that consume MCP services from other servers
• Run Clojure code through LLM tools in Claude Code or similar clients
• Integrate Clojure projects with the MCP ecosystem for tool and resource sharing
• Test MCP interactions locally using the in-memory transport

FAQ from mcp-clj?

What is the difference between mcp-clj and clojure-mcp?

mcp-clj offers simpler setup and self-contained evaluation