Sentinel is now listed in the official MCP Registry as io.rootstuff/sentinel. If you use Claude, ChatGPT, or any other MCP-compatible assistant, your monitoring is now something you can discover, connect, and operate from inside a conversation. This post covers what the registry is, why we published there, and how to connect your account in about two minutes.
What the MCP Registry Is
MCP (Model Context Protocol) is the open standard for connecting AI assistants to external tools. The official MCP Registry is its canonical index: a machine-readable catalog of servers that clients, directories, and agents use to discover what's available. Listings are namespace-verified, so io.rootstuff/sentinel can only be published by us, verified against the rootstuff.io domain. When an assistant looks for monitoring tools on your behalf, the registry is where it looks, and Sentinel is now there with a first-party, verified entry.
That last part matters. Most monitoring names you'd recognize either aren't in the registry at all or appear only as community-built wrappers around their APIs, maintained by third parties. Sentinel's listing is published and maintained by us, points at our hosted server, and needs nothing installed or self-hosted to use.
What You Can Actually Do With It
Sentinel's MCP server exposes 21 tools covering the whole product. In practice, that turns monitoring into questions and instructions:
- "Which of my client sites had downtime this week?"
- "List SSL certificates expiring in the next 30 days."
- "Create a monitor for newclient.com and check it from every region now."
- "Anything that needs my attention this morning?"
- "Pause the staging monitors while we deploy."
Reads cover uptime summaries, monitor details, incidents, performance and DNS history, expiring certificates, status pages, and on-demand global URL checks. Writes cover creating, updating, pausing, and deleting monitors, plus full incident management. If you manage client sites, the morning routine of clicking through dashboards becomes one question with a roundup answer.
How to Connect
In Claude Code, one command:
claude mcp add --transport http sentinel https://sentinel.rootstuff.io/mcp/sentinel \
--header "Authorization: Bearer YOUR_API_TOKEN"
Create the token under API tokens in your account, with only the permissions you want the assistant to have.
In the Claude and ChatGPT apps, add a custom connector pointing at https://sentinel.rootstuff.io/mcp/sentinel-oauth. You'll sign in to Sentinel and approve access through the normal OAuth flow, with no token to paste. Any other MCP client works the same way: OAuth endpoint for clients with a sign-in flow, bearer endpoint for headless ones. Full instructions, including per-client walkthroughs, live in the MCP docs.
Guardrails, Because an Agent With Delete Access Deserves Them
Giving an assistant control of production monitoring should not mean giving it the keys to everything:
- Access is scoped to your current team, and API tokens carry per-token permissions. A read-only token gives the assistant a read-only view, and a refused write names the missing permission instead of failing silently.
- Read tools are available on any plan with API access (Starter and up). Write tools require full API access (Pro or Business). See pricing for the breakdown.
- Deletes are irreversible and the server says so to the model, which in practice makes assistants confirm before destructive actions.
Why We Built This
We think the dashboard is becoming one interface among several, not the interface. Agencies and freelancers already live in chat with an assistant for half their workday; monitoring that can't answer questions there is monitoring that gets checked less often. The registry listing is a small milestone in a larger direction: Sentinel as monitoring that agents can operate end to end, from "add this client" to "write up last night's incident."
The server's public home, including connection details and the registry manifest, is on GitHub. If you build agents and want a tool added or changed, the issue tracker is open.
Try It
Connect your account with the steps above, or start from the MCP overview page for the full walkthrough. If you're not on Sentinel yet, see how the monitoring itself works first; the assistant integration comes along for free once your monitors exist.