Coding Tools MCP Server
File operations, git management, and command execution for AI coding agents.
Data last scanned yesterday · Reviewed 3 weeks ago
Overview
Coding Tools MCP provides a model-neutral runtime for managing codebases through the Model Context Protocol. It equips AI assistants with tools for file reading, search, multi-file patching, and interactive sessions, while maintaining strict workspace confinement.
Our verdict
With a security score of 73/100, this server is classified as medium risk. While the project includes features like Landlock kernel-level filesystem confinement and supports read-only modes to limit destructive actions, it remains a community-maintained tool rather than an official vendor release. Users should be aware that it is not yet editorially reviewed, so caution is advised when granting it access to sensitive environments.
- Supports Landlock for kernel-level filesystem confinement on Linux
- Zero direct dependencies minimize the software supply chain surface
- Provides deterministic, paginated results optimized for context windows
- Community-maintained with no official vendor backing
- Not yet undergone formal editorial security review
Tools
| Tool | Description | Risk |
|---|---|---|
| server_info | server, workspace, automatic project context, policy, runtime, auth, protocol, and fixed-catalog metadata. | low |
| check_exec_environment | lightweight execution policy and Landlock status. | low |
| get_default_cwd | inspect this MCP runtime's relative-path base. | low |
| set_default_cwd | change this MCP runtime's relative-path base. | medium |
| read_file | stream a bounded UTF-8 range without loading the whole file. | low |
| list_dir | list immediate or bounded-recursive directory entries. | low |
| list_files | iterate files with glob, ignore, hidden-file, sort, and cap controls. | low |
| search_text | literal or regex search; ripgrep stops after the result cap. | low |
| apply_patch | stage and atomically commit add/update/delete/move envelopes. | medium |
| exec_command | run a bounded command and wait up to 10 seconds by default. | high |
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| ChatGPT | ||||
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf |
Frequently asked questions
›What are the system requirements for this server?
The server requires Python 3.11 or higher and can be run using either the Python toolchain via uvx or the Node toolchain via npx.
›How does the server ensure file system safety?
The server restricts access to a single workspace root and explicitly rejects directory traversal attempts using '..' or symlink escapes. Additionally, on Linux systems, it utilizes Landlock for kernel-level filesystem confinement.
›Does the server support non-standard networking protocols?
Yes, while it defaults to --stdio, you can drop that flag to enable communication over Streamable HTTP on http://127.0.0.1:8765/mcp.
›How are security permissions managed?
Permission modes are used to gate specific capabilities such as network access, shell expansion, inline scripts, and destructive commands.
›Can I run the server in a containerized environment?
Yes, you can build a Docker image using the provided instructions to create a disposable sandbox environment, which is useful for working with untrusted code.
›Does the server provide a graphical interface?
Yes, by installing the 'desktop' extra, you can use a GUI version that supports per-workspace profiles, credential management, and live health checks.
Changelog
Badge
Maintain this server? Add the live badge to your README.