shippedit MCP server

Put your launch data inside your AI coding agent. One endpoint, one API key — works with Cursor, Claude Code, Codex, Windsurf, Zed and any MCP-compatible client.

shippedit runs a Model Context Protocol server over streamable HTTP. Once connected, your agent can read your products' launch scores, the failing audits behind them, the AI fix cards and the bottleneck analysis — and ask shippedit to run a fresh scan after you ship a fix. The server is read-only by default: it can never edit your listing.

Quick start

Create an API key.Dashboard → AI agent → create a key with the read scope (add rescan if you want the agent to trigger scans). The key is shown once — copy it immediately.
Point your client at the endpoint.https://shippedit.me/api/mcp
Authenticate. Every request carries your key as a bearer token: Authorization: Bearer sk_ag_…

Setup per client

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "shippedit": {
      "url": "https://shippedit.me/api/mcp",
      "headers": {
        "Authorization": "Bearer sk_ag_YOUR_KEY"
      }
    }
  }
}

Claude Code

claude mcp add --transport http shippedit https://shippedit.me/api/mcp \
  --header "Authorization: Bearer sk_ag_YOUR_KEY"

Windsurf / Zed / other HTTP clients

The same JSON shape most HTTP-transport clients use:

{
  "mcpServers": {
    "shippedit": {
      "url": "https://shippedit.me/api/mcp",
      "headers": {
        "Authorization": "Bearer sk_ag_YOUR_KEY"
      }
    }
  }
}

Tools

Six tools. All except trigger_rescan are read-only and free to call.

Tool	Scope	What it returns
list_products	any key	Your products on shippedit with their latest scan scores — the source of product keys for every other tool.
get_insights	read	Full latest report for one product: performance/AEO/authority scores, failing audits, the recommended First Edit, and unresolved AI fix cards.
get_fix_cards	read	Every stored AI fix card for one product — what/why/steps plus a paste-ready agent prompt per issue.
get_scan_history	read	Score-over-time series (last 100 scans): performance, AEO, OpenPageRank, with per-run source and timestamp.
get_bottleneck	read	The stored AI bottleneck analysis — the one edit that matters most, why now, expected impact, steps.
trigger_rescan	rescan	Starts a fresh full analysis in the background and returns immediately; poll get_insights until scanPending is false. Spends the same plan allowances as a dashboard rescan.

Prefer plain REST?

The same key authenticates a small REST API — useful for scripts and cron jobs:

Method	Route	Purpose
GET	/api/agent/v1/products	Your listed products with latest scores.
GET	/api/agent/v1/products/{key}	Full insights snapshot for one product + an agent-actionable summary (First Edit, fix cards, bottleneck).
POST	/api/agent/v1/products/{key}/rescan	Start a full rescan (rescan scope, 10 requests/minute per key). Returns remaining allowances + poll instructions.

curl -H "Authorization: Bearer sk_ag_YOUR_KEY" \
  https://shippedit.me/api/agent/v1/products

Keys, scopes and budgets

read — every reading tool. Unlimited calls; costs nothing.
rescan — unlocks trigger_rescan. A rescan spends the same plan allowances as the dashboard button (one full analysis + one page speed check for the current billing period) and is rate-limited to 10 requests per minute per key.
Keys can be restricted to specific products, given an expiry date, and revoked instantly from the AI agent page.

What this unlocks

"Check my shippedit scores, find the biggest issue, and fix it in this repo." — the agent reads get_insights, opens the First Edit's fix card, and implements it.
"Did last week's performance work move the needle?" — get_scan_history gives the before/after.
"Ship the fix, then rescan my site." — with the rescan scope, the agent ships the change and starts a fresh analysis in one session, then polls for the new numbers.

FAQ

What is the shippedit MCP server?

A Model Context Protocol endpoint that lets AI coding agents — Cursor, Claude Code, Codex, Windsurf, Zed — read your shippedit product scores, failing audits, AI fix cards, scan history and bottleneck analysis, and start full rescans. One endpoint, one API key.

How do I connect Cursor or Claude Code?

Create an API key in your dashboard (AI agent page), then add shippedit as an HTTP MCP server at the endpoint above with an Authorization: Bearer header. Exact configs are in the Setup section.

Is it free?

Yes — agent keys are free on every plan. Read tools are unlimited; trigger_rescan spends the same plan allowances as clicking Rescan site in your dashboard.

Can the agent change my listing?

No. The server is read-only by default and cannot edit listings, copy, pricing or images. The only write-like action is trigger_rescan, which re-runs an analysis and requires a key created with the rescan scope.

How do I revoke access?

Delete the key on the AI agent page in your dashboard — it stops working immediately. Keys can also be restricted to specific products and given an expiry date at creation.