Agent log
Watch every command your AI agent runs against your cases, live, with the kind of answer each one got.
#What it shows
The Agent log lists every command your AI agent runs through the discovery command-line tool, across every case, newest first. It's how you see your agent working, and how you check afterwards what it looked at.
Only the kind of answer each command got is shown, never the answer itself. The app doesn't keep the agent's answers.
Open Agent log in the left column.
#The columns
| Column | What it shows |
|---|---|
| Time | When the command ran (hover for the full date and time). |
| Case | Which case it ran on. |
| Command | The command and a summary of its arguments, for example search "wire transfer" --from 2019-01-01. |
| Answer | A badge for the outcome, plus a short result such as a number of hits. |
| Took | How long it took. |
#The outcome badges
| Badge | Meaning |
|---|---|
| ok | It worked. |
| not found | The document, citation or name it asked for isn't in the case. |
| invalid command | The agent typed the command wrong. It usually corrects itself. |
| needs you | A person has to do this step, such as approving highlights. |
| blocked by privacy | Refused, for example because agent access is off for that case. |
| busy | The case was being changed (a scan or a run). The agent should wait and retry. |
| engine unavailable | The app's engine wasn't reachable, for example because the app was closed. |
| error | Something went wrong. Details go to the app's local log, not to the agent. |
| over budget | Belongs to the app's own AI, which is switched off in this version. |
#Filtering and paging
- The Case menu at the top narrows the list to one case, or All cases.
- Beside it, Live shows the list is updating (it checks every two seconds), with last call 12 s ago and so on. If it says Not updating: the engine is not answering, the app's engine has stopped; restart the app.
- Load older at the bottom adds earlier calls.
#When it's empty
No calls yet means your agent hasn't run a discovery command since the log began. When it does, the call shows up here within a couple of seconds. If you expected calls:
- Check the agent was given the prompt from Start Up (Go to Start Up).
- Check the command-line tool is installed and current (the Start Up checklist).
- Check the agent app is allowed to run commands on this computer.
#Using the log well
- Watch the first few minutes of a new agent session. You'll see it read the guide, run
doctorandstatus, then search. It's reassuring, and it shows quickly if something's wrong. - Many
invalid commandrows in a row usually mean the agent didn't read the guide. Ask it to rundiscovery agent-guideand follow it, or write the guide files from Start Up. needs youis your cue: the agent should have told you what's needed.- Check its thoroughness. A claim that "nothing in the files mentions X" should sit on top of several searches with different wording. The log shows whether it did them.