Your AI assistant finally knows what you benched last Tuesday

What it does

hevy-mcp is a Model Context Protocol server for the Hevy fitness tracking app. It exposes 23 tools that let an AI assistant read your workouts, routines, exercise templates, and body measurements — and create or update them. So you can ask “analyze my last six weeks of training and find the gaps” in plain language and get an answer grounded in your actual log, not vibes.

The interesting bit

The architecture is more deliberate than most MCP glue: the repo is split into a web-safe Hevy client, an Effect-based operations layer handling retries, per-attempt timeouts, and interruption, and separate adapters for Node stdio, a Cloudflare Worker, and a standalone CLI. There’s also a hosted endpoint, so you can connect without running anything locally — the Worker validates your API key against Hevy per request and doesn’t store it.

Key highlights

• 23 MCP tools covering reads, mutations, and composite workflows like multi-week training summaries
• Runs hosted on Cloudflare (Streamable HTTP with OAuth) or locally via stdio, npx, bunx, or Docker
• Companion @chrisdoc/hevy-cli for terminal use, with --json output for scripting; deletions deliberately unsupported
• Compact, AI-friendly search results for routines and exercise templates to avoid blowing up the context window
• Claude plugin and workout skill included

Caveats

• Requires a Hevy PRO subscription — API keys are a paid feature
• The hosted endpoint requires a client that supports Streamable HTTP and a fixed Authorization header; otherwise you’re running the local stdio server
• The Effect internals are deep — fine for users, but contributors will want to read the runtime-scope docs first

Verdict

If you track workouts in Hevy and use an MCP-capable assistant, this is one of the more polished fitness integrations around. If you don’t pay for Hevy PRO, none of it works.