Coding Tools MCP
A model-neutral MCP server that gives any AI chat client or agent safe access to a codebase, with file editing, command execution, git, and permission-gated sandboxing.
Coding Tools MCP is a coding runtime served over the Model Context Protocol that turns any MCP-capable chat client, such as Claude Desktop, into a coding agent with real repository access, using whatever subscription the client already has. It exposes a default catalog of 18 tools for file reading and search, structured multi-file patches, command execution, interactive sessions, and git, confined to one workspace root per server.
Safety is enforced structurally rather than left to the model: absolute paths, .. traversal, and symlink escapes are rejected, permission modes gate network access, shell expansion, inline scripts and destructive commands, and Landlock adds kernel-level filesystem confinement on Linux. Results are summarized, paginated, and size-capped to fit context windows. It is vendor-neutral across Claude Desktop, Claude Code, Codex, Cursor, Cline, VS Code, Windsurf, and Gemini CLI.
It runs as a Python package from PyPI (with an npm launcher) over stdio or Streamable HTTP, and is released under the Apache-2.0 license.