Setup
Set up Observer with your coding agent in two minutes.
Paste one prompt and approve one link. Your agent gets a scoped key for the Observer MCP server, then builds your status page from the metrics you describe.
Terminal, Claude Code
> Set up Observer for me. Instructions: https://use.observer/llms.txt
Open this link to approve, and check the code matches:
https://use.observer/connect?code=WDJB-MJHT&client=claude-code
Code: WDJB-MJHT
Waiting for approval…
Approved. Connected to the acme organisation.
Continue here over the REST API, or restart Claude Code to load the Observer MCP tools?
Browser, use.observer/connect
Approve an AI agent connection
Approving creates an API key the requesting agent uses through the Observer MCP server. You can revoke it any time under Settings, API keys.
Sign in
Add the server and sign in.
Add the server with no key. Your agent opens Observer in your browser; you choose the organisation and approve the access it asks for.
Add the MCP server
Approval
An organisation owner approves. Each sign-in is backed by an API key, so it shows, and can be revoked, under Settings, API keys.
No OAuth?
If your client can't sign in this way, use the prompt or connect with a key below.
With the prompt
Or let your agent do the setup.
It reads Observer's instructions for agents, asks you to approve one link, and receives the key itself.
Paste into your agent
Approve
Open the link, check the code.
Your agent shows a link and a short code. Sign up or sign in, choose or create the organisation, check the code matches and approve. On the Free plan, approving starts a 14 day Starter trial once, with no card, because API access starts on Starter.
Connect
The key goes to your agent, once.
Your agent receives a scoped API key. It then asks whether to carry on in the same session over the REST API, or restart to load the Observer MCP tools. Restarting ends the session, so it asks before creating anything.
Describe
Say what your customers should see.
Name the services and the metrics behind them. The agent writes the config, shows the dry run, and applies it when you agree.
Run the agent
One command, for your environment.
Checks other than heartbeats run on an Observer Agent inside your network. The agent creates it and hands you the install command for Docker, Kubernetes, systemd or a binary. You run it; the key is shown once.
By hand
Or connect with a key.
Approve on the connect page, then add the MCP server to your agent's user config with your key in it, never to a file in a repository. The command below is for Claude Code; other agents take the same server URL and key.
Add the MCP server
Add the Observer skills
The key
The connect page shows the key once, already placed in the command. Store it in a secret manager if you keep a copy.
The skills
The skills teach your agent the setup loop: export, edit, apply. The command installs them for Claude Code; other agents can read the skill files from the repo.
github.com/useobserver/skillsAccess
What your agent can do.
Create and update status pages, services, metrics and SLOs, create the agents that run your checks, and read their status and incidents.
Granted
Set up and manage your status pages
- read:configExport your configuration and read your status pages
- write:configCreate, update and delete status pages, services, metrics and SLOs
- read:agentsRead agents and whether they are online
- write:agentsCreate agents, issue new agent keys and delete agents
- read:servicesRead services
- read:metricsRead metrics and their status
- read:slosRead SLOs and error budgets
- read:incidentsRead incidents
Not granted
Incidents, status overrides and SLA statements.
Writing incidents and maintenance, setting metric status, change events and SLA statements are left out. Config changes can still change what a public page shows, so the agent shows the dry run before applying. An owner can add the rest under Settings, API keys.
Revoking
Any time.
The key is listed under Settings, API keys, like any other, and can be revoked there at any time.