Logs & Run History

Every rule set run produces a run record and a stream of timestamped log entries, tied together by a correlation id so you can trace exactly what happened.

Runs and logs

Each environment keeps two kinds of records:

• Run — one entry per run, with its status (Pending, Running, then a final state), the rule set and project, who or what started it, timing, and any top-level error.
• Log entry — a single line written during the run. Each entry carries a timestamp, a severity level, the correlation id, and breadcrumb fields identifying where in the rule set it came from.

The Initiated By column names whoever started the run. A person is shown by name; an account that has since been erased reads as Deleted user rather than disappearing from the history. A run started by something other than a person names the credential kind instead — an API key, an MCP token, or the platform itself for a scheduled run. A run resumed from a human intervention names both the action that paused and the person who answered it.

Log levels

Log entries use standard severity levels:

The viewer shows Information and above by default. A debug logs toggle pulls in Trace and Debug entries on demand, so verbose diagnostics never clutter the normal view.

Correlation IDs

A correlation id is created when a run starts and stamped onto the run and every log entry it produces. When one rule set hands off to another — for example through a workflow or a continued run — the work stays grouped under a shared correlation, so a multi-step run reads as a single, ordered story.

Log messages can also reference other correlations. The viewer renders these as clickable links: selecting one opens that correlation's full log, letting you follow a chain across rule sets without losing your place.

Breadcrumbs

Each log entry includes breadcrumb fields that pinpoint its origin within the rule set: the rule set name, the rule, the action set, and the individual action (with its action type). The viewer renders these as a compact trail next to the message, so you can see at a glance which rule and action wrote a given line — and jump straight to the rule set in the editor.

The log viewer

The in-app log viewer lists recent runs newest-first and lets you expand any run to load its log entries inline. From the toolbar you can filter by project, and deep links can pre-select a specific project, correlation, run, or log entry.

• Live updates — while a run is Pending or Running, the viewer refreshes on its own so you can watch progress without reloading.
• Incremental loading — runs and log entries load more on demand, so large histories don't all load at once.
• Attachments — entries that captured files show download links you can click to retrieve them.
• By correlation — a dedicated dialog shows a single correlation's complete log, independent of the run list, and is what inline correlation links open.

Inference traces

Logs answer “what happened”; an inference trace answers “why.” When a rule set runs in Inference mode it can record a trace — a firing-by-firing account of the run. Turn it on with the rule set's Record run traces option (test runs are always traced); expanding such a run in the viewer then shows an Inference trace card alongside its log.

The trace viewer lays out:

• A halt banner and totals — whether the run reached a stable state, hit the activation cap, was cancelled, or errored, and how many firings and iterations it took.
• A timeline of firings: for each, the rule, its salience, the keys that triggered it, an expandable list of the context changes it made (old → new), and a “why it matched” condition tree showing the actual value at each clause.
• Derived facts resolved during the run, and events such as suppressed re-fires.

From a trace you can Explain with AI: pick a provider and model, optionally type a question, and get a plain-language explanation of why the run produced its result. Traces are size-capped, so a very long run may be marked truncated.

Recent logs

A compact “recent logs” panel surfaces the latest handful of runs (with status) for an environment — useful as an at-a-glance health check on a dashboard. Selecting a run takes you into the full viewer for detail.

Across many runs

Logs and traces explain one run. To see the pattern across thousands of them — run volume and success rate by day, how often each rule fires, and which decisions your rule sets produce — use the environment's Analytics page.