coding-mcp is a production-ready MCP server designed for one primary purpose: let AI agents write and ship code on remote repositories without ever cloning them to the agent machine. Deploy the server once on the machine that holds your code. Any MCP-compatible agent — Claude, Cursor, or any HTTP-capable client — connects to it and immediately has access to file operations, git, search, build, and test tools across all indexed projects. 
┌─────────────────────────────────────────────┐
│  AI Agent (Claude / Cursor / any MCP client) │
└────────────────┬────────────────────────────┘
                 │  MCP over HTTP or STDIO
                 ▼
┌─────────────────────────────────────────────┐
│           coding-mcp server                  │
│  (runs on the machine with your repos)       │
│                                              │
│  Tools: files · git · search · commands      │
│  Resources: project index · file content     │
│  Prompts: implement · review · fix tests     │
└──────────────────┬──────────────────────────┘
                   │  filesystem + git
                   ▼
        /projects/repo-a  /projects/repo-b  …

Quickstart

Install, configure, and connect your first agent.

MCP interface

Full reference for tools, resources, and prompts.

Source code analysis

Learn service boundaries and request flow.

Development

Build, test, and validate changes in the repository.

​
Core capabilities

CapabilityDescription
Multi-root project registryScan one or more root directories and persist a project index that agents query by stable ID
File operationsRead, write, create, delete, move files and apply unified diffs or structured edits
Git automationStatus, diff, branch, commit, push, tag — all with structured JSON output
Search and analysisFilename search, regex content grep, project tree, and stack summarization
Safe command executionBuild, test, and lint runners auto-detected from project tooling; arbitrary commands via an allowlist
Binary-safe resourcesFiles served as text or base64 blob depending on MIME type
HTTP auth and RBACAPI-key authentication with viewer, editor, and admin roles
OpenTelemetry hooksSpans for every HTTP request and tool operation
  1. Start with Quickstart to install and connect an agent
  2. Configure roots and limits in Configuration
  3. Secure HTTP deployments in Security and RBAC
  4. Browse the full tool reference in MCP interface