Docs

Your log, in whatever agent you already use.

Physique keeps your food, workouts and weigh-ins in plain JSON behind a small authenticated API. Point Claude Code, Codex, Pi or any tool with a shell at it and ask real questions about your own data.

Why you'd do that

• Ask about your week without opening the app. "What did my protein average last week?" "When did I last bench 225?" "Compare this cut to last summer's." The agent reads the numbers you logged; nothing is copied by hand.
• Log from wherever you're already typing. With a write credential an agent can add a meal or a weigh-in from a chat, a Slack thread, or a script. Save still goes through the same revision check the app uses, so nothing gets clobbered.
• Own the data. Export every day and every saved food as JSON in one call. Build a spreadsheet, a dashboard, or a model. Physique is the record; what you do with the record is yours.
• Safe by default. Credentials are read-only unless you choose otherwise, expire, are shown once, and can be revoked in the app. The server keeps only a hash.

How it works

1. In the app, open your account (the circle at the top right) and choose Agent access. Create a credential. Copy it now; it is never shown again.
2. Put the credential somewhere your agent can read it safely: a password manager, an environment variable, a secrets file with tight permissions. Never paste it into a chat.
3. Pick a route: the Claude Code skill, the MCP server (Claude Code, Codex, Claude Desktop, Cursor), the tiny Python CLI for Pi or anything with a shell, or the HTTP API directly.

What an agent can see