Autonomous Review Agent
Automatically request the best reviewers on high and critical risk pull requests using AI-driven selection from review history.
Overview
The Autonomous Review Agent monitors every pull request risk score computed by Koalr's Deploy Risk engine. When a PR is scored HIGH (70+) or CRITICAL (85+), the agent:
- Finds the most qualified reviewers from your org's recent review activity
- Requests those reviewers directly on the GitHub pull request
- Sends a Slack notification explaining who was requested and why
- Records the full assignment in the Autonomous Reviews dashboard
This closes the loop between risk detection and action — without requiring engineers to manually route PRs.
How to enable
The Autonomous Review Agent is controlled by an org-level feature flag. A Koalr staff member can enable it for your organization from the admin portal:
- Navigate to Admin → Feature Flags
- Find
autonomous_reviewer_enabledand toggle it on for your organization
Once enabled, the agent fires automatically on every PR scored HIGH or CRITICAL — no additional configuration required.
How reviewers are selected
The agent uses the following priority order to select up to 2 reviewers:
1. File-overlap matching (primary)
The agent looks at all PRs opened in the same repository within the past 30 days that touched at least one of the same files as the current PR. Reviewers who have reviewed the most of those overlapping PRs are ranked highest.
This surfaces engineers with direct, recent expertise in the specific files being changed.
2. Org-wide activity fallback
If no overlapping file reviewers are found, the agent falls back to the most active reviewers across all repositories in the organization over the past 30 days.
In both cases, reviewers who are already requested on the PR are excluded from consideration.
What data it uses
| Signal | Source |
|---|---|
| Changed file names | GitHub API (fetched at PR open/synchronize via webhook) |
| Review history | PullRequestReview records synced from GitHub |
| Feature flag status | FeatureFlag table (autonomous_reviewer_enabled) |
| GitHub token | Active GITHUB integration for the organization |
| Slack token & channel | Active SLACK integration settingsJson.botToken and settingsJson.defaultChannelId |
Viewing assignments
All reviewer assignments are visible in the Autonomous Reviews page in the Koalr dashboard (/autonomous-reviews). Each row shows:
- PR number and repository
- Risk level badge (HIGH / CRITICAL) and score
- Reviewers that were requested
- Reasoning (which signals led to their selection)
- Whether a Slack notification was sent
- Timestamp
Requirements
- GitHub integration must be active and have an access token with
pull_requests: writescope - Slack integration must be active with a
botTokenanddefaultChannelIdconfigured in settings (for Slack notifications — assignments are still recorded if Slack is not configured) - Feature flag
autonomous_reviewer_enabledmust be enabled for the organization
Weekly Risk Briefing
AI-generated weekly risk narrative delivered to engineering leaders every Monday — high-risk PRs, coverage declines, recent incidents, and reviewer expertise gaps summarized in plain English.
AI Tool Adoption Tracking
Track real AI coding assistant usage across GitHub Copilot, Cursor, and Claude Code — not just licenses, but actual activity, acceptance rates, and delivery impact.