Clojure

The Clojure MCP server connects AI models, like the ones from Anthropic or OpenAI, directly to a Clojure nREPL. It also bundles a set of specialized Clojure editing tools.

Features

• Clojure REPL Connection: Directly hooks into your running nREPL.
• Clojure-Aware Editing: Uses tools like clj-kondo, parinfer, cljfmt, and clj-rewrite for smart editing.
• Optimized Clojure Toolset: A comprehensive set of tools for Clojure development, even more than what Claude Code offers.
• Emacs Edit Highlighting: This is an alpha feature, but shows promise for Emacs users.

How to Use Clojure MCP.

Prerequisites

• Clojure (1.11 or later)
• Java (JDK 11 or later)
• Claude Desktop (recommended for the best experience)

Setting up the project

1. Get the Clojure MCP Server

git clone https://github.com/bhauman/clojure-mcp.git

Or, add it as a git dependency in your deps.edn:

{:deps {com.bhauman/clojure-mcp {:git/url "https://github.com/bhauman/clojure-mcp.git"
                                 :git/sha "latest-main-branch-sha"}}} ; Make sure to use an actual SHA

2. In the Clojure project where you want AI assistance, you need to update its deps.edn:

{:aliases {
  ;; nREPL server for AI to connect to
  :nrepl {:extra-paths ["test"]
          :extra-deps {nrepl/nrepl {:mvn/version "1.3.1"}
                       ch.qos.logback/logback-classic {:mvn/version "1.4.14"}}
          :jvm-opts ["-Djdk.attach.allowAttachSelf"]
          :main-opts ["-m" "nrepl.cmdline" "--port" "7888"]}
  ;; MCP server configuration
  :mcp {:extra-deps {org.slf4j/slf4j-nop {:mvn/version "2.0.16"}
                     com.bhauman/clojure-mcp {:local/root "/path/to/your/cloned/clojure-mcp"}} ; Or use the git dep
        :exec-fn clojure-mcp.main/start-mcp-server
        :exec-args {:port 7888}}}}

Remember to replace /path/to/your/cloned/clojure-mcp if you cloned it locally and are using :local/root. If you used the git dep, that part will be different.

A key point: the MCP server doesn’t have to run inside your project directory. It uses the nREPL connection for context. The root directory of the project running the nREPL server becomes the root for all MCP tool actions. Currently, the nREPL and MCP server need to be on the same machine due to an assumption of a shared file system.

3. You’ll need to edit your Claude Desktop configuration file. On a Mac, this is typically at ~/Library/Application Support/Claude/claude_desktop_config.json.

{
    "mcpServers": {
        "clojure-mcp": {
            "command": "/bin/sh",
            "args": [
                "-c",
                "cd /path/to/your/clojure/project && PATH=/your/bin-or-nix/path:$PATH && clojure -X:mcp :port 7888"
            ]
        }
    }
}

Adjust /path/to/your/clojure/project to your actual project’s path, and /your/bin-or-nix/path to your system’s binary path (e.g., /Users/yourname/.nix-profile/bin or wherever your clojure command lives).

4. Test the Setup

1. Start the nREPL in your target Clojure project:
   bash cd /path/to/your/project clojure -M:nrepl
   You should see a message like nREPL server started on port 7888....
2. Restart Claude Desktop. This is necessary for it to pick up the configuration changes.
3. Verify the connection. In Claude Desktop, click the + button in the chat area. You should see an option like “Add from clojure-mcp”.

5. In Claude Desktop, click the + (tools) button. You might want to add:

• Resource PROJECT_SUMMARY.md: The LLM can help create this. More on this below.
• Resource Clojure Project Info: This introspects the project connected via nREPL.
• Resource LLM_CODE_STYLE.md: Your personal coding style guide. You can copy the example from the clojure-mcp repository.
• Prompt clojure_repl_system_prompt: This provides instructions to the AI on how to code in Clojure.

Then, start your chat. You can begin by stating a problem and interactively design a solution with the LLM. Ask it to “propose” a solution. Iterate a bit. Then, you can have it:

• Code and validate the idea directly in the REPL. (LLMs are surprisingly good at this!)
• Make changes to your source code files and then validate in the REPL.
• Run tests.
• Commit the changes (make a branch first!).

Project Summary Management

The system has a neat workflow for maintaining a PROJECT_SUMMARY.md file. This file helps the AI quickly grasp your codebase structure.

• Creating/Updating: Use the MCP prompt create-project-summary (found in the + > clojure-mcp menu). It analyzes your code, documents key files, dependencies, and tools.
• Using: When you start a new chat, the “Project Summary” resource (if added) loads this file, giving the AI immediate context.
• Keeping Current: After a productive session where you’ve added new things, run create-project-summary again to update it.

FAQs

Q: What’s the main difference between using Clojure MCP and just, say, Claude Code directly with a Clojure project?
A: Clojure MCP offers a more deeply integrated experience with the Clojure REPL and a set of Clojure-aware editing tools. It’s designed as a cohesive system for REPL-driven development with AI, rather than just general code assistance. The stateful file tracking and specialized structural editing tools are key differentiators.

Q: Do I need to pay for LLM API keys to use Clojure MCP?
A: No. The core functionality, especially when used with Claude Desktop, does not require you to provide your own LLM API keys. The API keys are only necessary if you want to use the specific “agent tools” (dispatch_agent, architect, code_critique) which can make their own calls to models from Google, OpenAI, or Anthropic.