Docs

CLI and MCP server

One npm package. From a terminal it runs the site check, records deploys and lists what is up; from Cursor or Claude Code it is the tool the agent calls to create the monitors for what it just shipped.

Install and first run

npx serverinspector check example.com          # no account, no token

export SERVERINSPECTOR_TOKEN=si_...            # Settings, API tokens
npx serverinspector me
npx serverinspector monitors

Node 18 or newer. npm i -g serverinspector if you would rather not type npx.

Commands

command	does	token
check <hostname>	The free site check: online, response time, SSL, domain, DNS, redirects, security headers, page size, scored out of 100.	none
me	Who the token belongs to and what it may do.	any
monitors	List monitors with status, uptime and last response time.	read
incidents [--status ongoing]	List incidents, ongoing first.	read
create <name> <url> [--interval 60] [--keyword text]	A monitor; its first check runs at once.	write
run <monitor>	Check now. Exits 2 when the monitor is down, so a script can act on it.	write
deploy --monitor <name> [--version v] [--no-silence]	A deploy marker on the timeline, quiet for a few minutes while it restarts.	deploy
silence --monitor <name> --duration 10m	A maintenance window; lift ends it early.	deploy
mcp	An MCP server over stdio, for Cursor, Claude Code and anything else that speaks MCP.	read or write

--json prints the API's answer as it came; --api <url> points at another instance.

Token scopes

Made under Settings, API tokens. The secret is shown once. Whatever the scope, a token never reaches notification channels, billing, the account or the admin panel.

deploy: deploy markers and silences, by monitor name. What a CI runner holds.
read: list monitors, incidents and checks. What a dashboard or a read-only agent holds.
write: read, plus create monitors, run checks and deploy. What an agent that sets things up holds.

In CI

- run: npx serverinspector deploy --monitor api --version ${{ github.sha }} || true
  env:
    SERVERINSPECTOR_TOKEN: ${{ secrets.SERVERINSPECTOR_TOKEN }}

The || true is deliberate: if ServerInspector is unreachable, your deploy still goes out. The marker shows on the monitor's chart next to the response times, which is how "it got slow after Tuesday's deploy" becomes a fact.

As an MCP server

Claude Code:

claude mcp add serverinspector -e SERVERINSPECTOR_TOKEN=si_... -- npx -y serverinspector mcp

Cursor, in .cursor/mcp.json (Windsurf and others take the same shape):

{
  "mcpServers": {
    "serverinspector": {
      "command": "npx",
      "args": ["-y", "serverinspector", "mcp"],
      "env": { "SERVERINSPECTOR_TOKEN": "si_..." }
    }
  }
}

Tools: site_check, list_monitors, create_monitor, run_check, list_incidents, record_deploy. Then "monitor what I just deployed" is a sentence the agent can act on.

Source and issues

The package lives in the ServerInspector repository under packages/cli. Free plan, no card, to get a token.

Create an account