← All features

AI · Free+

AI agents for monitoring and incidents

Three agents inside StatusPulse do the tedious parts of monitoring: setting it up, working out what broke, and letting your AI assistant run it through MCP. You review everything they propose.

The Setup Agent: from your domain to a status page · 42 s
An incident, start to finish · 45 s
A real Claude Code session with the StatusPulse MCP server · 1:13

Monitoring is easy to buy and slow to set up: someone has to list every endpoint, certificate and domain worth watching, and during an outage someone has to read the graphs before anyone can tell customers what's going on. The three agents in StatusPulse take those jobs. They work inside your account, with the same plan limits as the dashboard, and they propose — you decide.

Setup Agent: from a domain to a status page

Type your domain. In about 20 seconds the Setup Agent looks at what anyone on the internet can see: your DNS records (including mail), your certificates, the usual subdomains such as app, api and docs, and health endpoints like /health. It proposes the probes worth having — website, API, web app, certificates, domain registration, mail routing — names and groups them, and decides which belong on your public status page. Claude reviews the list first; it can rename, group and drop probes, but never invent addresses the scan didn't find.

You switch off what you don't want and click once: the page, its component groups and the probes are created. No port scans, nothing behind a login. On every plan, including Free: 3 scans a day on Free, 5 on Starter, 20 on Pro and 50 on Business.

Incident copilot: the probable cause, while it's happening

On every incident, the copilot reads all the probes on the status page and the timeline from 30 minutes before the incident opened. Because StatusPulse checks databases, gRPC services, WebSockets and email — not just URLs — it can say "the Postgres probe slowed down 90 seconds before the API started returning 503s" instead of restating the error. You get up to three probable causes, each with the observations behind it and the next thing to check, a drafted public update, and who on call to page first.

On incidents StatusPulse opens automatically, it runs the moment you open them. Nothing is posted and nobody is paged until you do it. Pro includes 10 analyses a day, Business 100.

Run it from your AI assistant: the MCP server

Connect Claude Code, Cursor, VS Code or any other MCP client and run your monitoring in plain words from where you already work: put a new service on your status page, ask what's down, open an incident and keep customers updated while you fix it. Your assistant does it through the StatusPulse MCP server; a REST API with the same tools sits beside it, and the instructions an agent needs are at statuspulse.ai/agent.

Ask it to…Tools it uses
Put a new service on a status pagecreate_status_page, create_probe
Set up monitoring for a new domainscan_domain, apply_setup
Tell you what's down, with the last errorget_status, list_probes
Tell customers about an outage, then resolve itopen_incident, post_incident_update
Find the probable cause (Pro and Business)analyze_incident
Review or tidy uplist_status_pages, list_incidents, delete_probe

Opening an incident or posting an update reaches your subscribers, so those tools tell the assistant to show you the text and wait for your yes; deleting a probe asks first too.

The quickest way in is one click: under Pages → API tokens, Create setup prompt gives you a prompt with a new agent token in it. Paste it into Claude Code, Cursor, VS Code or another assistant that can add MCP servers, and it adds StatusPulse to its own configuration, checks the connection and offers to set up your monitoring. Or add the server yourself:

claude mcp add --transport http statuspulse https://statuspulse.ai/mcp \
  --header "Authorization: Bearer sp_agent_…"

Only Owners and Admins can create an agent token. It acts in one workspace with Editor rights, every change it makes is in the audit log, and revoking it stops it at once. On every plan.

Your data and the AI

The Setup Agent's review and the copilot use Anthropic's Claude, listed as a sub-processor in our privacy policy. They send only what the job needs — the scan of your domain, or the incident's probe results and the display names of the people on call — never credentials, request headers or e-mail addresses. An Owner or Admin can switch every AI feature off for the workspace; the Setup Agent then still works, with its built-in rules.

Plans: Free+ · Help: AI agents for monitoring and incidents

Guides

Related

Try AI agents for monitoring and incidents