What Bots Monitoring shows
Every review bot in the workspace, what it flags, how often people act on it, and what it costs.
On this page
Most teams now have several bots commenting on their pull requests: AI reviewers such as CodeRabbit, Copilot or Greptile, quality gates such as SonarCloud, and dependency bots such as Dependabot. Bots Monitoring is where you find out which of them earn their place. It measures what each review bot says, how serious it is, how often your team acts on it, and what it costs, for every repository in the workspace.
#The tabs
- ROI: one row per bot, with its threads, how many were used, how many were left, the severity mix, the cost per used thread and a verdict: Keep, Tune or Rarely used.
- Themes: what the bots keep raising, grouped into themes by AI. Shown only while AI is on.
- Benchmark: how each bot does for you against the same bot in other teams’ repositories of a similar size.
Press 1, 2 or 3 to switch tabs. A repository’s own page has a Bots tab with the same three views, measured over that repository alone (see Repository pages).
#What the bots are flagging
On the ROI tab, a summary strip counts the bots’ findings in the window: how many there were, the share that is high severity (major plus critical), the share of nits and the top topic. Each bot’s row opens its breakdown. The picture below is a working model with made-up data.
| Bot | Threads | Used | Not addressed | Severity mix | High |
|---|---|---|---|---|---|
| 132 | 58% | 55 | 44% | ||
| 125 | 61% | 49 | 54% | ||
| 41 | 80% | 8 | 51% | ||
| 41 | 59% | 17 | 44% |
The severity and topic come from a small model that runs on your computer, not from the bot’s own label. See Severity scoring.
#The two callouts
Only a bot reviewed
When open pull requests have had a review from a bot and none from a person, a callout says how many. Select Show list → to see them. Consider a human pass before they merge.
Likely-addressed bot threads
When later commits look as if they dealt with bot threads that are still open, a second callout counts them. Select Show list → to review them and resolve them on GitHub in one go. “Looks addressed” is a guess (a commit touched the file after the comment), so check each one before you resolve it.
#Which accounts count as bots
Limn recognises most bots by their GitHub account and how they behave, and gives each one a category: Review bot, Quality check, Dependency updates, Code agent, Release automation or Housekeeping. Only review bots are measured on the ROI table; quality checks are listed apart, because nobody is meant to reply to them. To change a bot’s category, mark an account as a bot or a person, or set its price, open Settings → Bots.
#What you can do
- Choose Bots Monitoring in the sidebar.
- Pick a window on the ROI tab: 7d, 14d, 30d or Sprint. The other tabs use the same window.
- Select a bot’s name to open its comments and the pull requests it touched.
- Select Depth → on a bot’s row for its response times, coverage, consistency and severity over time.
A bot’s activity, comment by comment, is on the Feed’s Bots side. A single pull request’s bots are on its Bot activity tab.
#Why it helps
A bot that comments on everything trains people to ignore it, and its useful comments get lost. Bots Monitoring shows you the signal-to-noise of each one, in numbers you can check comment by comment, so you can tune or drop the noisy ones and keep the ones your team acts on.