Skip to main content

​
Coding Tools MCP

Coding Tools MCP exposes practical coding operations as a Model Context Protocol server. It gives an MCP client a controlled way to inspect a workspace, search text, apply patches, run commands, interact with long-running processes, inspect Git state, request permissions, and view local images. The project is not an agent framework. It is the tool layer a client can connect to when it wants code-aware capabilities inside a bounded workspace.

​
What it provides

  • Workspace file access through read_file, list_dir, list_files, and search_text.
  • Code editing through apply_patch.
  • Process control through exec_command, write_stdin, and kill_session.
  • Git inspection through git_status, git_diff, git_log, git_show, and git_blame.
  • Runtime metadata through server_info, check_exec_environment, get_default_cwd, and set_default_cwd.
  • Permission escalation flow through request_permissions.
  • Image inspection through view_image when enabled.

​
Runtime model

The package installs the coding-tools-mcp command. By default it serves Streamable HTTP on 127.0.0.1:8000/mcp. Passing --stdio switches it to newline-delimited JSON-RPC over stdio. The implementation advertises MCP protocol version 2025-06-18 and supports initialize, ping, logging/setLevel, tools/list, and tools/call.

​
Safety model

Coding Tools MCP is designed around a workspace root. Direct file tools resolve paths inside that workspace, follow explicit size limits, and reject unsafe or unsupported inputs. exec_command is controlled separately by permission modes: Tool exposure can also be reduced with --tool-profile read-only or adapted for clients with --tool-profile compat-readonly-all.

​
Start here

Use Quickstart to launch the server, Installation & Configuration to choose transport and security settings, Client integration for MCP client configuration, and Tools reference when you need exact tool names and argument behavior.