MCP setup guide

Setup in four steps

1. 01

   Create a Liftw account

   Sign up free and log a few sessions so your training summary has real data to reason about.

   Get started
2. 02

   Generate an API key

   Open Progress → AI Analysis and create a key. Copy it once — Liftw stores only a hash. You can revoke it anytime.

   Open Progress
3. 03

   Connect your MCP client

   Point Cursor, Claude, or another Streamable HTTP MCP client at the production endpoint with your Bearer key.

4. 04

   Ask for analysis

   Have the agent pull your summary or coach brief, reason about the last weeks, then publish an insight back to your dashboard.

Production endpoint

Health check: https://mcp.liftw.xyz/. The Worker proxies authenticated tools to https://www.liftw.xyz.

Client configuration

Use Streamable HTTP / remote MCP where your client supports it. Replace the placeholder with the key from Progress. Exact config shape can differ slightly between Cursor and Claude Desktop.

{
  "mcpServers": {
    "liftw": {
      "url": "https://mcp.liftw.xyz/mcp",
      "headers": {
        "Authorization": "Bearer lfw_YOUR_KEY_HERE"
      }
    }
  }
}

Available tools

• get_training_summaryFull snapshot — progress, recent workouts, profile slice.
• coach_briefCompact text brief optimized for the model context window.
• list_insightsInsights already posted to your AI Analysis panel.
• publish_insightWrite a new analysis to your private dashboard.

Suggested agent flow

1. 1Call coach_brief or get_training_summary
2. 2Reason about volume, PRs, imbalances, and recovery
3. 3Call publish_insight with a clear title and summary
4. 4Open Liftw Progress → AI Analysis to review the post

Example prompt: “Summarize my last 4 weeks of training and post the analysis to Liftw.”

Privacy

• No public profiles and no social feed.
• MCP access exists only while your key is active.
• Revoke keys anytime from Progress → AI Analysis.
• Training logs are treated as health-related personal data.

See the Privacy Policy for full details.