Agent activity
Every AI run in the project: which agent, what it was about, each step it took, and what people did with the result.
Where Sidebar → Agent activity, or How was this made? under an AI answer
What you see
- Agent filterShow All agents or one: Extractor, Critic, Research, Continuity, Ask, Rule check, Images or Transcribe.
- RunsNewest first. Each row shows the agent, when it ran, what it was about, the outcome, the number of AI steps and tool calls, time and tokens. Failed runs are marked Failed with the error.
- Run detailsOpen a row to see the model id and prompt version (for provenance), the effort, what the run was for with a link to it, and what happened Afterwards.
- StepsA table of every step: what ran (the AI, a web search or a tool), its input, its result, time and tokens. A continuity run also shows its report.
See how an answer was made
- Under an AI answer, a research note or a continuity report, choose How was this made?.
- Agent activity opens on that one run with its steps expanded.
- Choose All runs to go back to the full list.
Check what happened to extracted candidates
- Choose Extractor or Critic in the agent filter.
- Read the outcome on each row: how many candidates were accepted, rejected or are still waiting in Review.
- Open a row and follow the link to the document it read.
Advice, not decisions
The runs here only proposed or answered. Nothing an agent did became canon on its own; the Afterwards line shows what people decided. A continuity check of a private draft is shown only to its author.
Who can do what
- Every member
- Read every run in the project (except continuity checks of other people’s drafts).
Roles are per studio. Roles and permissions has the full list.