# Futureweb Intervals MCP [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/futureweb/intervals-mcp-server  
**GitHub Stars:** 5  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/futureweb-intervals-mcp

## Description
Self-hosted Intervals.icu MCP server: training analytics, Garmin data, wellness, workout planning

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "futureweb-intervals-mcp": {
    "command": "uvx",
    "args": ["futureweb-intervals-mcp@1.0.0b1"]
  }
}
```

## Documentation

## What Futureweb Intervals MCP does

Futureweb Intervals MCP MCP server exposes Intervals.icu data to AI assistants through MCP. It is designed for athletes and developers who want an agent to inspect training history, interpret endurance performance, compare planned and completed workouts, or prepare workouts from information stored in Intervals.icu.

The server reads activities with custom fields and streams, intervals, wellness records, thresholds, zones, calendar events, climbs, best efforts, nutrition, weight, and training summaries. It also supports analysis of recovery, fatigue resistance, efficiency, aerobic durability, intensity distribution, and projected training load. Device-specific values remain separate from Intervals.icu load values.

## How it works

The application runs in the environment chosen by the operator and communicates with the Intervals.icu API. It does not contact Garmin directly. If the optional Garmin Intervals Bridge has restored values such as training effect, recovery time, VO₂max, stamina, sleep, HRV, or running dynamics into Intervals.icu, Futureweb Intervals MCP can read those fields and streams dynamically.

MCP clients can connect locally or remotely. The server supports stdio, streamable HTTP at `/mcp`, and SSE from one process. Remote access includes an OAuth 2.1 server, an Intervals.icu sign-in flow, consent pages with per-connection permissions, PKCE, RFC 9207, and client metadata documents using `private_key_jwt` for ChatGPT connections.

## Setup and configuration

Futureweb Intervals MCP MCP server is self-hosted and requires access to an Intervals.icu account and API. Configuration determines which permission classes are available. Read permissions are enabled by default; write, destructive, and admin permissions must be enabled separately.

Heavy tools accept `detail_level` values of `compact`, `standard`, or `full`, and may return JSON through `output_format="json"`. The project also provides MCP resources containing coaching prompts and method guides, with guide retrieval through `get_guide`.

## Tools and capabilities

Available capabilities include:

- Reading activity details, custom fields, streams, intervals, wellness data, calendars, and training summaries.
- Returning streams as summaries, CSV, or JSON, with slicing, paging, downsampling, and per-interval statistics.
- Comparing planned and executed workout steps, including deleted events and activity extensions beyond the original plan.
- Analyzing climbs, descents, best efforts, repeated workouts, power meters, efficiency, fatigue resistance, and aerobic durability.
- Producing recovery snapshots using 42-day baselines, wellness trends and correlations, nutrition summaries, and weight trends.
- Creating and validating workouts when write access is enabled, including route-aware steps that can end on a lap press.
- Showing the exact write request with `dry_run`, refusing a duplicate event on the same day, and reading back stored events after a write.

The tool catalogue supports compact read and write configurations, which can reduce the context required by an MCP client.

## Limitations and notes

The project is marked as public beta and is not affiliated with Intervals.icu or Garmin. Garmin enrichment depends on the companion bridge or another system that writes the relevant custom items into Intervals.icu; the MCP itself never contacts Garmin.

Write operations are not available by default and depend on permission configuration. Interpretations are left to the connected coaching agent rather than being presented as fixed medical or training advice. The server is intended for data held in Intervals.icu, so it is not a direct Garmin Connect integration.

_Full upstream README: https://allmcps.com/mcp/futureweb-intervals-mcp/readme_