Model Context Protocol

Connect your AI assistant to your monitoring

Sentinel speaks the Model Context Protocol, so Claude, ChatGPT and any MCP-compatible agent can read your monitoring and act on it. Ask “is anything down?”, triage incidents and pause noisy monitors, all in plain language.

Is anything down right now?

Yes, 1 of 12 monitors is down. api.acme.com has been failing for 4 minutes (HTTP 502 from 3 of 4 regions). Everything else is operational.

Pause it so it stops paging, and open an incident.

pause_monitor · create_incident

Done. Paused api.acme.com and opened incident #1487 “502 from origin.” I'll leave your other 11 monitors running.

Why MCP?

Answers, not dashboards

Most mornings you don't want a dashboard, you want an answer. With Sentinel connected to your assistant, you ask and it reads the result straight from your account, then takes the next step for you when you approve it.

Ask questions in plain language instead of clicking through dashboards
Let agents triage incidents, pause noisy monitors and run one-off checks
Standard protocol, so it works with the AI tools you already use

MCP tools

Ways to connect

Regions for live checks

24/7

On-demand access

Features

Everything your assistant can reach

Answers in plain language

Ask “is anything down?” or “what needs my attention this morning?” and your assistant reads the answer straight from your account. No dashboard digging.

21 built-in tools

Uptime summaries, an attention roundup, monitor and incident inspection, performance and DNS history, certificate expiry, and on-demand global URL checks.

Read and write by plan

Read tools work on any plan with API access. Write tools (create, update, pause monitors and manage incidents) require full API access on Pro or Business.

Team-scoped and safe

An assistant only ever sees the team you're currently working in. Every write asks for your approval first, and deletes are flagged as destructive.

Works with any MCP client

It's standard Model Context Protocol, so it works with Claude, ChatGPT, Cursor, your own agents and IDE extensions. No lock-in.

OAuth or bearer token

One-click OAuth sign-in for hosted apps like Claude and ChatGPT, or an API bearer token for Claude Code, Claude Desktop and custom agents.

Instructions for agents

Connect in under a minute

Point your assistant at one of two endpoints. Hosted apps use OAuth; developer tools and custom agents send a bearer token. Both expose the same tools, scoped to your current team.

OAuth endpoint

Claude & ChatGPT apps

One-click sign-in, no token to copy.

Bearer token endpoint

Code & custom agents

Send an API token as an Authorization header.

Claude Code

claude mcp add --transport http sentinel https://sentinel.rootstuff.io/mcp/sentinel --header "Authorization: Bearer YOUR_TOKEN"

Claude Desktop & custom MCP clients

{
  "mcpServers": {
    "sentinel": {
      "type": "http",
      "url": "https://sentinel.rootstuff.io/mcp/sentinel",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

Replace YOUR_TOKEN with an API token from your account settings. Full setup, including ChatGPT and other clients, lives in the MCP docs and the MCP tools reference.

FAQ

MCP questions

What is MCP?

The Model Context Protocol is an open standard for connecting AI assistants to external tools and data. Sentinel ships a built-in MCP server, so any MCP-compatible assistant can read your monitoring and act on it in plain language.

Which AI assistants can I connect?

Any MCP-compatible client. That includes the Claude apps and Claude Code, ChatGPT on paid plans, Cursor and other IDE extensions, plus your own agents. Hosted apps connect over OAuth; developer tools use a bearer token.

Is it safe to give an assistant access to my monitoring?

Yes. The connection is scoped to the single team you're working in, read tools are all it needs for routine questions, and every write asks you to approve it first. Deletes are explicitly flagged as destructive, and usage is tracked alongside your other API activity.

Do I need a paid plan?

Any plan with API access can use the read tools, asking “is anything down?” or “which sites are slowest this week?”. The write tools that create, update, pause monitors or manage incidents require full API access on Pro or Business.

What can the assistant actually do?

Read tools cover uptime summaries, an attention roundup, listing and inspecting monitors and incidents, performance and DNS history, expiring certificates, and on-demand URL checks from every region. Write tools create, update, pause and resume monitors, run one-off checks, and acknowledge, resolve or update incidents.

How do I connect it?

Paste the OAuth URL into a hosted app like Claude or ChatGPT, or run one command in Claude Code with a bearer token. Full step-by-step setup lives in the docs, and the connection snippets are on this page.

Bring your own assistant

Ask your monitoring anything

Connect Claude, ChatGPT or your own agent and get answers in plain language. 14-day free trial on paid plans, cancel anytime.

Start Free Trial → Read the MCP docs

Works with any MCP client

OAuth or bearer token

Team-scoped access