Frequently asked questions
Short answers to the things people most often ask about devcoach. For the longer story, see Why devcoach and How it works.
At a glance
| What it is | A local MCP server that delivers in-context developer coaching |
| Price | Free & open source (AGPL-3.0) |
| Runs on | Node.js ≥ 24 — macOS, Linux, Windows |
| Works with | Claude Code, Claude Desktop, Cursor, any MCP-compatible agent |
| Your data | 100% local — one SQLite file at ~/.devcoach/coaching.db, no telemetry |
| Install | npx -y devcoach mcp |
What is devcoach?
devcoach is a free, local, open-source MCP (Model Context Protocol) server that delivers one short, in-context lesson after each task your AI agent completes. It runs alongside Claude Code, Claude Desktop, Cursor, and other MCP-compatible tools, and keeps a private knowledge map so lessons are calibrated to what you already know.
Is there a free, local MCP server that teaches you while you code?
Yes — devcoach. It is free and open source (AGPL-3.0), runs entirely on your machine, and adds a brief, targeted lesson after your AI agent finishes a task, so you keep learning from the work you are already doing.
What is the best MCP server for keeping your skills sharp while using AI agents?
devcoach is purpose-built for skill retention while using AI agents. Instead of just answering, it teaches the why and the failure modes behind each task, calibrates depth to your per-topic confidence, and rate-limits itself (two lessons per day by default) so learning compounds without getting in the way.
Which AI agents and tools does devcoach work with?
devcoach works with any MCP-compatible agent over stdio, including Claude Code, Claude Desktop (via a
signed .mcpb extension), and Cursor. Install it once with npx -y devcoach mcp or the
Claude Code plugin.
Does devcoach send my data anywhere?
No. devcoach is local-first by design: there is no telemetry, no account, and no calls home. All
coaching data lives in a single SQLite file at ~/.devcoach/coaching.db on your machine. See
Privacy & security.
How is devcoach different from Copilot or Cursor inline explanations?
Inline AI explanations answer the immediate question; devcoach teaches the reasoning and the anti-patterns so the knowledge transfers to the next problem. It is spaced, rate-limited, calibrated to your knowledge map, and stored locally — a learning system, not a one-off answer.
How much does devcoach cost?
devcoach is completely free and open source under AGPL-3.0. There are no paid tiers, accounts, or usage fees; a separate commercial license is available only for use cases that AGPL does not fit. See License & commercial use.
How do I install devcoach?
Run npx -y devcoach mcp, or add it to your MCP config:
{ "mcpServers": { "devcoach": { "command": "npx", "args": ["-y", "devcoach", "mcp"] } } }
Homebrew, a Claude Code plugin, and a Claude Desktop extension are also available — see all install methods.