MCP Server
Connect ChatGPT or another compatible Model Context Protocol client to Quantified Self, approve only the read-only data categories you want it to use, and ask questions about your training history without exposing your entire account by default.
Training Questions
The server exposes bounded tools for discovery and analysis. Every result is projected through the permissions you approve instead of returning raw account documents.
Discover available activity metrics, query bounded identity-free body-weight history by day, week, or month, and inspect ready Training analysis such as load, readiness, intensity, durability, and sport-specific evidence.
Discover canonical Sports Lib activity types, filter bounded newest-first scans for requests such as your latest run, resolve today or yesterday in an explicit IANA timezone, and inspect summaries, laps, swim lengths, MTB jumps, app links, selected numeric metrics, or bounded chart-ready data. Metric ranking supports explicit periods or a processing-bounded all-history scan, so a client can use the relevant persisted maximum-jump metric as the authoritative record and request individual jump details only when needed.
Ask for one bounded sleep trend that combines recorded-vital coverage with duration, score, stages, HRV, heart rate, blood oxygen, and respiration across daily, weekly, or monthly buckets. With both sleep and Training-metrics permission, ask for one daily report with your latest completed sleep, recorded aggregate HRV and sleep heart rate, Dashboard Today’s live UTC-day Readiness, and current-versus-usual 28-day Training totals and sport mix. It keeps readiness concise and exposes safe same-provider baselines and evidence states without returning provider identity, raw physiological samples, a workout plan, diagnosis, or medical advice.
Filter bounded newest-first route scans by canonical activity type or route-name text, then read names, metrics, counts, and timestamps. A separate saved-route location permission enables exact bounds, simplified geometry, nearby search, segment endpoints, and waypoints.
Access Boundaries
MCP access uses an authorization flow designed for external clients. The client receives only the scopes you approve and can be disconnected from Connections.
Activity and Training metrics, body measurements, individual activity details, activity locations, sleep summaries, saved-route summaries, and saved-route locations are explicit grants. Each location scope depends on its matching data scope, and the activity and route location domains remain independent.
MCP clients cannot create, edit, or delete activities, routes, sleep records, dashboard settings, or Training state. Revocation blocks future access.
Original files, full-resolution recordings, absolute sample timestamps, unrequested streams, provider credentials, source keys, and device provenance are excluded. Exact coordinates appear only under the matching location permission.
Direct-coordinate searches stay within Quantified Self. For a place-name search, only the supplied location text is sent to Mapbox for geocoding—not your activity, route, account, or unrelated prompt data.
FAQ
It gives a compatible external client a bounded, read-only way to discover and query approved Quantified Self activity metrics, body-weight history, Training analysis, individual activity details, sleep summaries, and saved routes.
Yes. Add https://quantified-self.io/mcp as a custom app endpoint in ChatGPT, sign in to Quantified Self, and approve the read-only permissions you want to grant. ChatGPT is an external client with its own privacy and retention practices.
No. The MCP server is read-only. It cannot write dashboard settings, activities, routes, Training state, or sleep records.
No. An activity chart may selectively parse an existing original file in memory, but the file itself, full-resolution recording, absolute sample timestamps, unrequested streams, provider payloads, credentials, and storage paths are not returned or copied into another activity store.
Only if you separately approve Activity locations or Saved-route locations and geometry. Activity location covers starts, ends, MTB jumps, nearby activity search, and optional breadcrumbs. Route location covers bounds, previews, segment endpoints, nearby route search, and waypoints. Granting one never exposes the other.
Next Step
Create an account, connect the training sources you use, and authorize only the MCP permissions needed for the questions you want to ask. Review or revoke completed client connections at any time.