Workbooks: runbooks and release checks
A workbook is a runnable workflow of monitoring and automation steps, drawn as a graph. Use one to check a release end to end, to diagnose a known failure the same way every time, or to hand a pass or fail result back to your CI pipeline.
- Node types
- Synthetic, observability, collectors, logic, actions, AI
- Runs
- On demand, on a schedule, or from CI
- Control flow
- Branches, loops, approvals, retries
- Authoring
- Designer, templates or Discovery
Building blocks
| Group | Nodes |
|---|---|
| Synthetic | Web test, API test |
| Observability | APM observe (read traces and service health) |
| Collectors | Database query, log collector, network probe, through agents inside your network |
| Logic | Assertion, branch, for each, approval, delay, wait for HTTP |
| Actions | HTTP action, plus PagerDuty, Opsgenie and Jira |
| AI | Analyse, probe, evaluate, eval suite, model compare |
What you can do with them
Gate a release
A pipeline triggers a workbook over the API and polls for a pass or fail result, so a deploy only continues when the checks pass. Results can also show as a GitHub check on the pull request.
DocumentationRepeatable diagnosis
Encode the steps an engineer would take for a known failure: check the endpoint, read the traces, query the database, then branch on what was found.
DocumentationHuman approval
Pause a run at an approval step before anything with side effects, such as a remediation call, goes ahead.
DocumentationSchedules
Run a workbook on a cron schedule as a recurring end-to-end check.
DocumentationCompare runs
Replay an execution and diff two runs to see which step changed.
DocumentationVariables and secrets
Pass outputs between steps with dot-paths, accept run-time parameters, and reference secrets without exposing them.
Documentation
Discovery: a first draft from a URL
Discovery opens your application in a real browser on a synthetic agent, records the API calls each page makes, and drafts a monitoring workbook from what it saw. You review and edit the draft before it runs.
Real browser
The crawl runs in a browser, so the calls captured are the ones your front end really makes.
Draft, not deploy
The output is a workbook you open in the designer and change as needed.
Templates
Or start from a ready-made workbook in the template gallery and point it at your system.
DocumentationLive timeline
Each run shows node status, retries and outputs as it goes.
Documentation
Read more in the docs
Try it on your own systems
Create a workspace, or tell us about your environment and we will walk you through it.