Agents

Coding agents like Claude Code, Cursor and Codex can add SignalSitter to your code or run it for you. They work about as well through any of three routes. Pick whichever your agent already handles.

CLI

The SignalSitter CLI is coming soon.

Until then, your agent can run curl against the REST API:

curl --fail-with-body \
  --header "Authorization: Bearer $SIGNALSITTER_API_KEY" \
  "https://api.signalsitter.com/v1/monitors"

See the REST API guide for more.

REST API

Your agent calls https://api.signalsitter.com directly, with the credential as a Bearer token.

• REST API guide: authentication, examples and errors.
• API reference: every operation.
• /openapi.json: the full contract in one file. Agents read it well.

MCP

If your agent supports MCP, connect it to https://api.signalsitter.com/mcp. It then sees SignalSitter’s tools and calls them by name. Tools that change something only run when the agent sends confirm: true.

Set up MCP has the config for Claude Code, Codex, VS Code and the TypeScript SDK.

Give your agent the docs

Point your agent at one of these:

• /llms.txt: an index of every page.
• /llms-full.txt: all the docs in one file.

Then ask for what you want. For example:

Read https://docs.signalsitter.com/llms-full.txt, then add a SignalSitter
heartbeat to the end of my nightly backup job.

Credentials

Create a credential in the dashboard. The secret appears once, so copy it straight into your agent’s environment as SIGNALSITTER_API_KEY. Never paste it into a prompt, a committed file or a URL.

Give the agent only what the job needs:

Preset	What it can do
write	Send beats. Reads nothing.
read	Read monitors, observations, incidents, callbacks and usage.
readwrite	Everything read can, plus send beats.
manage	Everything, including creating and configuring monitors, callbacks and credentials.

A job that only sends beats needs write. An agent that only looks things up needs read. An agent that sets up monitors or callbacks needs manage.

Give each agent its own credential. Then you can revoke one agent’s access in the dashboard without breaking your jobs.

Each credential works in one project. To work in another project, create a credential there.