AI agents
Let Claude Code, Codex, Cursor or another MCP client read and operate your monitoring.
Glowo speaks the Model Context Protocol, so an AI assistant can query your monitoring and act on it — with permissions you grant explicitly and can revoke at any time.
What an agent can do
Depending on the scopes you grant:
- Investigate — list monitors, read status, pull heartbeats and statistics, fetch an SLA report.
- Operate — create and update monitors, pause and resume, run a check now, acknowledge an alert.
- Communicate — open and update incidents, publish to a status page, schedule maintenance.
- Configure — manage notification channels and send a test.
In practice that means asking “which monitors are down and what changed recently?” and getting a real answer, or “pause the staging monitors for an hour” and having it happen.
Clients
Connecting has copy-paste setup for Claude Code, Codex CLI, Cursor and VS Code, plus the discovery URLs any other OAuth-capable MCP client needs.
Two ways to connect
| Hosted | Local | |
|---|---|---|
| Runs | On Glowo’s servers | On your machine |
| Auth | OAuth 2.1 with a consent screen | A gl_ API key |
| Setup | Paste a URL, approve in the browser | Build from source |
| Revoke | Per-connection, in the dashboard | Revoke the API key |
| Best for | Almost everyone | Self-hosted Glowo, or offline work |
Start with the hosted connection. It needs no install, grants scopes explicitly through a consent screen, and can be revoked per client.
Safety
- Scopes are explicit. An agent gets exactly what you approve, and read-only is a first-class option.
- Some things are never available. Billing, member management, workspace settings and the audit log cannot be granted to an agent or an API key.
- Everything is attributable. Agent actions appear in the audit log like any other.
- Revocation is immediate, per connection.
Start read-only. Grant write scopes once you have watched what the agent actually does with them.