MCP & AI

AI agents can operate Uptimely through a first-class Model Context Protocol surface — the same domain logic as the dashboard, behind the same authorization.

Connect your coding agent

Paste the copied line into Claude Code, Codex, Cursor or OpenCode and the agent fetches the setup prompt and configures itself: it adds the remote MCP server, and — where you want programmatic access too — walks through minting a project API key for the REST API and SDK. See the setup guides for the exact steps per client, including the ones without a one-shot prompt (claude.ai, ChatGPT).

Remote MCP server

Uptimely hosts a remote MCP endpoint at https://app.getuptimely.com/api/mcp (Streamable HTTP) with OAuth 2.1 authorization: connect it from any MCP-capable client (Claude, ChatGPT, and other agent hosts), approve the consent screen, and the agent gets tools over your projects — monitors, incidents, alerts, status pages and more.

Authorizations are visible and revocable under User Settings → Connected AI apps.

Write access is opt-in

Read tools work out of the box. Write operations are gated twice: the project-level "Allow AI write operations" toggle (under Settings → API Keys) must be on, and the /v1 write surface additionally requires the Pro plan. The same toggle governs the copilot's write tools.

In-app copilot

The dashboard ships an AI copilot (⌘I, or the launcher in the corner): ask it about your monitors and incidents, or let it do the clicking for you — every write it proposes requires your explicit in-chat approval before it runs.

API keys and agents

Project API keys (uptimely_live_…) also authenticate machine access — see API. OAuth-minted agent tokens and API keys are deliberately separate credential types: /v1 accepts project API keys, the MCP surface accepts OAuth tokens.