for MCP server makers

Glimind for Makers

Know the minute your MCP server breaks, what changed, and how agents are using it — from the people who already check it.

Maker
$12/month

One MCP server you own.

Studio
$29/month

Up to five MCP servers you own.

Start from your listing's manage page after you claim your server.

What you get

Down and recovered alerts

Email, Slack, Discord or a signed webhook within about a minute of our checks seeing your server go down, with the failure class, and again when it recovers with the outage length.

Checks every minute

Your server is handshaked every 60 seconds instead of every 5 minutes. Checks stay liveness-only: initialize and tools/list, never a tool call.

Tool-definition change alerts

A diff of exactly which tools were added, removed or changed whenever what you serve changes — the same record the public transparency log keeps.

Deploy attestation

Run glimind attest in CI to record the manifest you meant to ship. If what is served does not match within 15 minutes, you hear about it.

Regression alerts

Readiness drops of 10+ points, new tool-poisoning findings and client-compatibility regressions (OAuth metadata, protocol version, Origin checks) from a daily conformance run.

Incident updates on your status page

Post investigating / identified / resolved notes on glimind.com/status/<your server>, shown as your words next to our measurements.

Monthly report

Uptime, outages and time to recover, latency, readiness trend, tool changes and agent demand for the month, by email.

Agent demand

How often agents' tasks were routed to your tools, how often you were the runner-up, and the unmet tasks in your categories. Counts only — no task text is stored.

Written-for-you parameter descriptions

AI-drafted descriptions for your undocumented parameters, marked for you to check before you ship them (60 a month).

Instant re-audit

Re-probe and re-score right after you deploy a fix, as often as once a minute.

What stays free, for everyone

Your public score, grade and findings, the status page, the transparency log, corrections to your listing, the signed proof, the readiness audit with fixes, glimind check in CI and deploy attestations. These are the same for every server whether or not anyone pays.

Our promise

Paying never changes your score, grade, findings, ranking, routing or what the public log shows, and never removes anything. Every server is measured the same way whether or not its owner pays. A plan buys your private view of that data, alerts, reports and help fixing what we find.

How it works

  1. Find your server and claim it: put a token in your repo, your endpoint's /.well-known or a DNS TXT record. You get an owner token.
  2. Pick a plan on the manage page and add alert channels: an email, a Slack or Discord webhook, or any https endpoint (signed with HMAC).
  3. Add one line to CI so every deploy is checked against what goes live:
    pip install glimind
    GLIMIND_OWNER_TOKEN=${{ secrets.GLIMIND_OWNER_TOKEN }} glimind attest <your-server-id> --url http://localhost:8000/mcp

Checks are liveness only: the MCP handshake and tools/list, from one region. We never call your tools. Questions: [email protected]. Machine-readable: /v1/maker/plans.