mcp-claude-code
MCP implementation of Claude Code capabilities and more
claude mcp add --transport stdio sdglbl-mcp-claude-code node server.js \ --env MCP_LOG_LEVEL="info" \ --env MCP_ALLOWED_ROOTS="<comma-separated-directories-if-applicable>" \ --env MCP_MODEL_PROVIDER="<lite-llm-provider-config-if-applicable>"
How to use
This MCP server implements Claude Code-like capabilities, enabling Claude to inspect, modify, and manage a codebase through a rich set of file and project operations. It exposes tools for reading and writing files, performing precise text edits, and executing shell commands across supported languages. In addition to basic file I/O, the server provides advanced features such as recursive directory exploration, pattern search with ripgrep integration, AST-aware code search, notebook cell handling, and structured task/delegation workflows. Agents can be dispatched to work concurrently, allowing Claude to orchestrate complex multi-step tasks while maintaining proper permission and security checks.
How to install
Prerequisites:
- Node.js installed on your machine (version 14+ recommended)
- Access to the repository containing the MCP Claude Code server
Installation steps:
- Clone the repository: git clone <repository-url>
- Navigate to the project directory: cd mcp-claude-code
- Install dependencies: npm install
- Configure environment (examples):
- In a .env file or your environment, set: MCP_LOG_LEVEL=info MCP_ALLOWED_ROOTS=/workspace MCP_MODEL_PROVIDER=<your-lite-llm-provider-config>
- Start the MCP server: node server.js
- Verify the server is running by sending a test MCP request or visiting the health endpoint (as documented in INSTALL.md).
If your setup uses a container or a different runtime, adapt the start command accordingly (see mcp_config in this document).
Additional notes
Notes and tips:
- Ensure the server’s allowed roots and permissions match your project layout to avoid permission errors during file operations.
- When using read/write/edit-related tools, prefer atomic operations (e.g., multi_edit) for large or critical changes to minimize partial updates.
- For large codebases, enable batch or dispatch_agent workflows to parallelize tasks and improve responsiveness.
- If you integrate with an external LLM provider, validate input sanitation and output handling to prevent unintended file modifications.
- Review the Security section in the README for recommended practices and potential caveats related to shell command execution and file system access.
Related MCP Servers
pfsense
pfSense MCP Server enables security administrators to manage their pfSense firewalls using natural language through AI assistants like Claude Desktop. Simply ask "Show me blocked IPs" or "Run a PCI compliance check" instead of navigating complex interfaces. Supports REST/XML-RPC/SSH connections, and includes built-in complian
rlm-claude
Recursive Language Models for Claude Code - Infinite memory solution inspired by MIT CSAIL paper
stt -linux
Local speech-to-text MCP server for Tmux on Linux (for use not only with Claude Code)
3dsmax
3dsmax mcp server for Claude and other AI agents.
CogniLayer
Persistent memory for Claude Code & Codex CLI — save ~100K tokens/session. 13 MCP tools, hybrid search, TUI dashboard, crash recovery. Your AI finally remembers.
powerpoint
MCP server for creating and editing PowerPoint presentations