Week of October 3–9, 2026
Public API 🔥
LitmusLab now has a public, documented API: trigger suite and collection runs, fetch results, and read monitor data programmatically, using the same endpoints the dashboard itself is built on. Authenticate with an Org API key, managed under Org Settings → API Keys, and you're off. Access CI pipelines, internal scripts, anything that needs to drive LitmusLab without a browser.
The full reference is generated straight from the API's own spec, so it's never out of sync with what's actually deployed. See the Public API docs to get started.
MCP Server 🔥
You can now connect Claude directly to your LitmusLab projects. Add it as a custom connector in Claude.ai, sign in once, and ask Claude things like "what's failing in the Checkout suite" or "trigger the Nightly collection and tell me how it goes"; it's calling the same API above, scoped to exactly the projects you're a member of, not your whole org like an API key.
Nothing to generate for this one — it authenticates through your normal LitmusLab login. See the MCP Server docs for setup.
Smarter pass/fail for API Monitoring
API Monitor results used to be a strict and over every check; one slow response could sink an otherwise-correct evaluation with no way to override it. Now, if your suite has an Intent set, a failing check gets a second, holistic look from the same judgment LitmusLab already uses for Test Suite runs, adjusted for the fact that a Monitor capture is a partial, live-traffic walk rather than a full synthetic run. Checks are now grouped by step in Recent Runs too, so you see what actually happened along the way instead of one flat list.
Fixes worth knowing about
API Monitor's Overview and Sessions tabs no longer duplicate each other — Sessions now carries the full per-capture detail (response, pass/fail, step-grouped checks, Intent Evaluation), and Overview shows aggregate tiles instead, including a new breakdown for captures that never produced a run at all because they matched no step or hit one configured to bypass checks. A redundant 0%/100% "pass rate" label and a Failure Breakdown table that could only ever show the same two numbers are both gone. The editor's tabs are renamed and reordered to match UI Monitoring's own convention (Overview, Config, Sessions), tab clicks now actually refetch instead of showing stale data, and hovering a run's response shows the full text instead of a second truncation on top of an already-cut-off one. Elsewhere: the left nav's Projects section now collapses, suites get a small icon badge for UI vs API workflow type, and UI Monitoring's per-monitor name field autosaves instead of needing a Save click.