About Paceva
Paceva is a read-only MCP server for endurance athletes, turning personal watch data synced through Intervals.icu into readiness and race-prediction information. Predictions are ranges checked against the athlete’s history, with device and Paceva values identified separately.
Key features
- 01Readiness and training data access
- 02HRV and sleep data access
- 03Workout history analysis
- 04Race predictions as personal ranges
- 05OAuth 2.1 sign-in
Use cases
- 01Assess whether to train hard today
- 02Estimate a current swim race time
- 03Review a recent cycling training block
MCP support
Exposes an MCP server at `https://paceva.app/api/mcp` using Streamable HTTP and OAuth 2.1 with PKCE. Tools provide read-only access to the authenticated user’s Paceva data.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://paceva.app/api/mcpclaude mcp add --transport http owenscheuer-paceva https://paceva.app/api/mcp{"mcpServers":{"owenscheuer-paceva":{"type":"http","url":"https://paceva.app/api/mcp"}}}Open in Cursor{"servers":{"owenscheuer-paceva":{"type":"http","url":"https://paceva.app/api/mcp"}}}Open in VS Code{"mcpServers":{"owenscheuer-paceva":{"type":"http","url":"https://paceva.app/api/mcp"}}}{"mcpServers":{"owenscheuer-paceva":{"serverUrl":"https://paceva.app/api/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project README
Paceva for developers
Paceva turns an endurance athlete's watch data into honest race predictions: ranges, checked against their own history. This repository documents the public surface: the Paceva MCP server, the OAuth 2.1 sign-in in front of it, and how Paceva connects to Intervals.icu.
The Paceva app itself is closed source. This repo is documentation and examples only.
Connect in one minute
Server URL: https://paceva.app/api/mcp
| App | How |
|---|---|
| Claude (claude.ai, desktop, mobile) | Settings → Connectors → Add custom connector, paste the URL, sign in to Paceva, approve. |
| ChatGPT | Settings → Apps & Connectors → add a custom connector with the URL (some plans need developer mode). |
| Claude Code | claude mcp add --transport http paceva https://paceva.app/api/mcp, then /mcp to sign in. |
| Cursor / VS Code | One-click buttons on paceva.app/dashboard/ai, or add it as a Streamable HTTP server. |
| Anything else | Any MCP client that supports Streamable HTTP and OAuth 2.1 with PKCE. |
Then ask things like "Should I go hard today?", "What could I swim a 1500 in right now?" or "How did my last block of riding go?".
Full walkthrough: docs/connect.md.
What's here
| Doc | For |
|---|---|
| docs/connect.md | Connecting each AI app, and what to expect |
| docs/tools.md | Every MCP tool, its arguments and what it returns |
| docs/oauth.md | The OAuth 2.1 server: discovery, PKCE, client registration, tokens |
| docs/intervals-icu.md | How watch data gets in: the Intervals.icu sign-in, step by step |
| docs/data-and-privacy.md | What the AI app can and can't see |
| docs/listing.md | The MCP Registry entry (server.json) and directories to submit to |
| examples/ | curl discovery, a raw JSON-RPC call, client config snippets |
Ground rules
- Read-only. One scope,
paceva:read. No tool writes anything. - Only your own data. A token is bound to one Paceva account; it can never read anyone else's.
- Every number is labelled.
devicevalues are your watch's own (e.g. Garmin's HRV);pacevavalues are computed by Paceva. - Predictions are ranges. Tools return a low–high range and how often that range holds for you. Quote the range, not a single time.
- Revocable. Disconnect any app from Paceva → Use with AI; its tokens stop working on the next call.
Status
| Piece | Status |
|---|---|
| MCP server + OAuth | Live |
| Intervals.icu sync (Garmin, Coros, Wahoo, Polar, Suunto) | Live |
| Paceva web dashboard | Coming soon, join the waitlist |
| iPhone app (Apple Health) | Coming soon |
Security
Found a problem? Please email [email protected] rather than opening a public issue. See SECURITY.md.
License
Docs and examples in this repository are MIT licensed. The Paceva name and logo are not.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.