Set up monitoring from your editor

One command. Your AI assistant does the rest.

Connect exit1.dev to Claude, Cursor, Codex or any MCP client. Your assistant reads the repo, works out what should be monitored, creates the checks, configures alerts, and fires a real test alert to prove delivery works. Then you can just ask it about uptime.

Get started freeSign In

Free forever · No credit card · Sign up with GitHub or Google

Paste this into Claude Code, then ask it to set up monitoring for your project:

claude mcp add --transport http exit1 https://app.exit1.dev/mcp/v1

OAuth in the browser — no API key. Sign-up happens inline if you don’t have an account. Other clients and the npm package.

Key Features

Everything you need to monitor your infrastructure effectively

It Sets Monitoring Up For You

Point your assistant at a project and it reads the repo — deploy configs, env files, health route handlers — works out what should be monitored, and creates the checks. You approve the list; it does the typing.

One Command, No API Key

The hosted server uses OAuth. Run one command, approve in the browser, done. Sign-up happens inline if you don't have an account yet — no dashboard visit, no key to paste, no config file to edit.

14 Tools — Read And Write

Seven read tools for uptime, history and stats. Seven write tools to create and update checks, pause them, set email recipients, and connect Slack, Discord or Teams. Delete is opt-in and never granted by default.

It Proves The Alert Works

The setup playbook finishes by firing a real test alert, because a channel that was configured but never tested is the most common way monitoring silently fails. You see the alert arrive before you walk away.

Real Health Checks, Not 200-Checking

Your assistant finds your /health endpoint and asserts on the response body with JSONPath — because a 200 with a dead database is still a 200. Plus DNS drift, domain expiry, TCP ports, ICMP and cron heartbeats.

Then Ask It Anything

"Is anything down?" "What's the uptime on my API this month?" "Show me the last ten failures." "Compare response times this week vs last." No dashboard, no context switch.

Why Choose exit1.dev?

See how we stack up against the competition

Featureexit1.devOthers
Agent creates and configures monitors for you
Hosted MCP server with OAuth (no API key)
Ships a setup playbook as an MCP prompt
Natural language monitoring queries
Claude, Cursor, VS Code, Windsurf, Codex, Gemini, Goose
Per-scope permissions, revocable per clientN/A
Listed on the official MCP registry
Historical data and multi-range stat comparison
Available on the free planEnterprise only

Technical Details

Built for developers, by developers

Architecture

Two transports, one identical tool surface. The hosted server is a stateless streamable-HTTP endpoint at app.exit1.dev/mcp/v1, authenticated with OAuth 2.1 (PKCE, dynamic client registration) — your client registers itself, so there is nothing to configure. The local option is the exit1-mcp npm package running on your machine over stdio with an API key. Every tool in both is a thin wrapper over the public REST API, so tier caps, scope checks, rate limits and URL validation are enforced server-side and cannot be bypassed through the MCP surface.

Performance

Most tool calls return in well under a second. The hosted server needs no install and no warm-up; the local package starts instantly via npx and carries a single runtime dependency, the official MCP SDK. Nothing is stored on your machine beyond your client's own config.

API

14 tools. Read: get_account, list_checks, get_check, get_check_history, get_check_stats, get_status_page, get_alert_settings. Write: create_check, update_check, toggle_check, delete_check, set_email_alerts, add_webhook_alert, send_test_alert. Plus a setup_monitoring prompt carrying the full playbook. Every tool is annotated with readOnlyHint, and delete_check with destructiveHint, so your client can warn you before it acts. Published as exit1-mcp on npm and as dev.exit1/exit1-mcp on the official MCP registry.

Frequently Asked Questions

Everything you need to know about our monitoring

MCP (Model Context Protocol) is an open standard that lets AI assistants connect to external tools and data. Exit1's MCP server gives your assistant both read and write access to your monitoring — so it can set monitoring up for you, then answer questions about it in plain language.

Run one command: claude mcp add --transport http exit1 https://app.exit1.dev/mcp/v1. Your assistant opens a browser, you sign in (or sign up — it happens inline), approve the permissions, and you're connected. No API key, no config file, no Node.js install.

Claude Code and Claude Desktop, Cursor, and VS Code with Copilot support the hosted OAuth server today. Windsurf, Codex CLI, Gemini CLI, Goose and ChatGPT work via the local npm package (exit1-mcp), which exposes exactly the same tools using an API key instead of OAuth. Remote MCP with OAuth is still rolling out across clients.

It can create, update and pause monitors, and configure alert channels — that is the point, and it is what lets an agent set monitoring up for you. Deleting is separate: the checks:delete scope is never granted unless a client explicitly asks for it, and the consent screen shows you every permission before you approve. You can revoke any connection at app.exit1.dev/mcp, which kills every token issued to that client immediately.

"Set up monitoring for this project." "Is anything down right now?" "What's the uptime for my API over 30 days?" "Show the last 10 failures for production." "Add a check on staging and send alerts to #ops in Slack." "Monitor my nightly billing cron as a heartbeat."

Yes. The hosted server has no plan requirement at all, and the local npm package needs an API key that every plan can mint, Free included. Normal plan limits still apply either way — Free covers 50 monitors at 5-minute intervals, with DNS and domain checks reserved for paid plans. Your assistant calls get_account first so it plans within whatever your plan allows.

Not for the hosted server — it runs on our infrastructure and your client connects over HTTP. The local option needs Node.js 18+ and installs itself via npx, with no manual package install.

The MCP server runs on the same public API rate limits as everything else, applied per key and per user. A typical assistant conversation uses a handful of tool calls, so you are unlikely to notice them. Current numbers are in the API reference in our docs.

API & Webhooks

Full REST API access for programmatic integration. The MCP server is built on top of the same API.

Analytics & Reports

Track uptime trends and response times. MCP lets you query the same data conversationally.

Smart Alerting

Email, SMS and webhooks. Your assistant can configure these for you — and prove they work.

Status Pages

Public status pages for your customers. Query their current state via MCP.

Live Checks

Watch probes stream live in the browser while your AI assistant queries the same data via MCP.

Global Monitoring

Monitor from multiple regions. MCP surfaces data from all regions in a single conversation.

Start Monitoring