Product changelog
Every release that moved the needle on PR risk scoring, policy enforcement, and incident governance for engineering teams running AI coding agents.
AI code review Check Runs that never finished now resolve cleanly. A background reaper closes stale in-progress reviews after 15 minutes so the PR's Checks tab stops showing a perpetual "Computing risk signals" spinner. When a new commit replaces an in-flight review, the older Check Run is closed as superseded instead of orphaned.
Clicking "View more details on Re-Entry.ai" from the GitHub Checks tab now opens the matching pull request's assessment panel directly, instead of dropping you on the dashboard list to hunt for it.
Every code review specialist now reads from a single shared change manifest assembled once per pull request — file contents, intent, repo conventions, and linked tickets. Reviews are faster, less duplicative, and less likely to flag a fix that's already scoped to a follow-up ticket.
A "Cancel AI review" button now appears on the in-progress Check Run while a review is running. Use it when a PR is opened by mistake or needs more work before review without force-pushing.
Reassessments no longer fire when a PR's only new commit is a merge of main. Reassess comments are also suppressed when the score change stays within the same severity band, cutting noisy "Reassessed: 33 → 35" updates that didn't change the verdict.
Changes to local-only files like .vscode/settings.json or .env.local are now capped at LOW risk and never block a merge. Idiomatic editor + local-env updates stop generating critical findings.
Comment `@re-entry-ai` (or `@reentry`) on any pull request to trigger a fresh AI review that bypasses the cache. Useful after editing the PR description, updating labels, or wanting a second opinion. The bot reacts with 👀 so you know it heard you.
Assessments now key off a hash of the canonical PR inputs (head commit, title, body, labels, base branch) instead of a 1-hour time window. Editing any of these triggers a fresh review against the new state; identical inputs return the cached verdict instantly.
The Check Run on every pull request is now named "re-entry · AI code reviewer" and includes a "View more details on re-entry.ai" link in the Checks tab summary so you can jump from GitHub to the full assessment in one click.
A new Mission Control toggle lets teams exclude pull requests whose title starts with "Revert" from assessment. Reverts authored by scripts and bots that can't act on PR comments no longer get reviewed when the toggle is on.
The AI code reviewer now drops findings that hedge with words like "Possible N+1" or "Possible XSS" unless the body cites concrete evidence — a real caller, route, attacker-reachable endpoint or runtime metric. Plausibility no longer counts as evidence, so reviews focus on substantiated risks.
Livestream PR cards now read reassess progress from server state instead of ephemeral session memory, so the two-step overlay stays accurate across tab changes and page refreshes until the AI review is actually ready.
The livestream reassess overlay now advances to step 2 once risk assessment finishes, even while the AI review is still generating, instead of staying on step 1 until the entire operation completes.
Reassessing a pull request now fills the livestream card with a two-step progress overlay and keeps it locked until the AI review is ready, with clear status text instead of a silent spinner.
Reassessing a pull request now regenerates the same code review the dashboard shows and updates the tracked GitHub review in place, instead of leaving stale bot comments behind.
Fix Agent auto-PRs now run generated patches through compile, lint, and test checks in Vercel Sandbox before opening a pull request. Unverified fixes stay as suggestions.
Open scan findings now group into ranked clusters on the findings page. Fix, suppress, or convert an entire cluster to a guard instead of working row by row.
The intelligence panel can now draft AGENTS.md from your repo's manifests, hooks, and CI config. Review, edit, adopt privately for reviews, or open a GitHub PR to commit conventions to the repo.
Deep scan status, cadence, and top open findings now live in the Risk Intelligence panel. The standalone scan page redirects to the dashboard so scan signal stays next to PR risk context.
PR reviews now reconcile your changes against open Heatshield scan findings on the same files and their blast radius. Reviews focus specialists on what the diff actually touches instead of running every agent on every PR.
Inline reviews now show your top 4 non-critical findings instead of the previous 5, matching the cap our benchmark testing showed produces the highest signal-to-noise. Every critical finding still posts. When findings are hidden, the review body says so transparently. Teams that want a different cap can override it from their team settings.
The follow-up comment posted when a PR is reassessed now shows the score you actually saw in the previous review on the left of the arrow, not the new score on both sides. Reassessments with no change render as 'Reassessed — no score change' instead of a misleading 'X → X' arrow.
Every PR now shows a readiness percentage and verdict — Blocked, Needs work, Almost there, or Ready to ship — in a chip beside the title. Click it to see which findings from the initial review are still open, which were resolved by which commit, and how the score has moved across each push. The engine tracks findings across pushes so fixes actually lower the score.
Draft pull requests no longer trigger a review until they're marked ready. A new team setting lets you opt out of auto-review on bot-authored PRs (Dependabot, Renovate, etc.) — useful when bot PRs are reviewed by your own workflow instead.
When Dependabot PRs include a changelog in the PR body, the engine now extracts and feeds it into the review so the LLM understands what actually changed in the bumped dependency rather than scoring it as opaque churn.
The guards panel has been rebuilt from the ground up. Each guard card now shows its condition summary, action pills, and integration health at a glance. New users see an interactive flow strip explaining how guards work, with three one-click starter templates. An 'Analyze my codebase' button scans your repo and returns 3–5 tailored guard suggestions powered by AI.
Connecting Sentry via OAuth now completes successfully. Guards that create tickets will now correctly route to your connected Linear or Jira account without requiring any additional configuration.
Commands that call the backend now show a live spinner while waiting. Medium-risk commits surface a visible warning instead of passing silently. A new REENTRY_TIMEOUT_MS env var lets you tune the 60-second request timeout. Logging out now warns when your token is still wired into Claude Code or Cursor.
The graph sidebar now has three switchable lenses — Blast Radius, Risk Map, and Architecture. Click any file to see its incident history, revert count, dependents, and risk score in a slide-in drawer. Risk driver chips filter the graph down to each driver's affected files so you can bridge from the assessment text to the visual instantly.
PR and push cards now show a GitHub / GitLab / Bitbucket badge so teams with multiple providers can tell signals apart at a glance. A new source filter narrows the feed to one provider. The repo selector in the dashboard header now aggregates repos across all connected providers instead of GitHub only.
GitLab MRs now show up in the open PRs list and pushes to monitored branches show up in the push feed. Until today the risk assessment was running but the results weren't being routed to the dashboard tables — only the underlying event log.
The sign-in modal now leads with three OAuth providers. Email/password is folded behind a single "Continue with email" button so the most common flow is one click, and returning email users still land on the form they used last time.
OAuth callbacks now verify a server-issued, single-use state value before granting access — preventing forged callback links from linking integrations to the wrong account. Applies to GitHub, GitLab, Bitbucket, Linear, Slack, Jira, Gmail, Calendar, and WorkOS SSO.
GitLab merge requests now flow through the full risk engine with file-level detail. The integration was passing an incomplete payload to the assessment pipeline, which caused assessments to short-circuit and occasionally pull data from the wrong PR. GitLab now produces assessments on the same signals as GitHub and Bitbucket.
Bitbucket pull requests now run through the full risk engine. The integration was missing file-level detail and refreshing access tokens, so most Bitbucket PRs were scoring as low risk regardless of what they changed. Bitbucket and GitHub assessments now operate on the same signals.
Every blog post ends with a "More from the blog" section showing the three most recent articles (current post excluded). Internal linking helps you discover related content faster — and tells search engines how our articles connect.
Content editors can now override meta title, meta description, keywords, canonical URL, noindex, OG image, and OG alt text per blog post directly in Storyblok — no code deploy needed. Defaults still come from the post itself when an override isn't set.
Canonical host is now uniformly https://www.re-entry.ai across every page, sitemap, robots, and structured-data block. The home page now ships a single <h1>, blog post bodies can no longer break the document outline from the CMS, and overlong meta descriptions on three top pages have been trimmed below Google's truncation limit.
When a CLI or MCP call is rejected (e.g. the repo isn't in your monitored set, or you've hit a plan limit), the dashboard's audit-log row now shows the actual reason and remediation inline — no more opaque "ERROR" entries. The full structured error code, message, and recommended fix ride along on the same row.
The codebase graph now reads the authoritative file list from GitHub for each PR instead of inferring it from heuristic signals. PRs that previously showed '0 changed' or '1 changed' now show the real number, with proper direct- and transitive-dependency tiers. The stale-index indicator is driven by an explicit index status record, not a wall-clock guess.
The codebase graph in PR and push assessments now switches between Blast Radius, Risk Map, and Structure views. Risk drivers double as tap-to-filter chips, and clicking a file opens a detail panel with dependents, incidents, reverts, change frequency, and a GitHub link.
Reports on credentials, backend reachability, git pre-commit hook, IDE agent configs (Claude Code + Cursor), and CLI version in one go. JSON envelope is CI-friendly: pipe it through jq to fail a build on any check that flips to `fail`. Bearer token is never printed.
The counterpart to `reentry init`. Removes the managed pre-commit hook and restores any backup we made on the first install. Refuses to delete a foreign hook (one we didn't write). Leaves credentials in place; use `reentry logout` to remove them.
Dashboard can now ask the API whether a compliance export will succeed without firing one off. While the export is gated pending legal sign-off, you get a clean badge with the reason instead of a 403 on click.
Breach-resolution errors now carry curated, user-safe messages instead of raw database error text. Schema details no longer leak into the dashboard when something fails.
Risk Intelligence now leads with the dollar value of incidents prevented over the last 30 days, calibrated against your team's own 90-day incident rate. When sample is too small, falls back to industry defaults — and tells you so on the tile.
GitHub, Copy, Reassess buttons and the right-edge chevron now hit a proper 44px tap target on phones. Desktop spacing is unchanged.
Each executed intervention now shows a quantified savings estimate. Click the pill for the full breakdown: incident cost prevented, re-burn token cost, dev hours, and the explicit assumption behind each number.
Interventions now appear indented directly under the change that triggered them, with a connector line. The activity stream reads as one conversation per change, not two parallel feeds.
Tightened input validation and rate limits on the compliance and savings endpoints. Cleaner error messages — no internal schema details leak when something goes wrong.
At the bottom of each assessment, tap a context-aware prompt ("What's the production impact?", "Why does this driver score so high?", "Suggest a guard that would have caught this") to open the chat panel pre-seeded with the review.
Archive, Intervene, Watch branch, and Fix actions now live in the full PR assessment. AI Act breaches with pinpoint + recommended fix surface at the top of every PR + push review.
PR + push reviews now show critical, high, and medium risk drivers by default. Low-severity findings sit behind a single "Show N" expander.
Hover the GitHub link on any PR or push card to see what it opens — same tooltip pattern as the other quick-action buttons.
Clicking a Pull Request or push card now opens the full assessment surface directly: side rail to navigate sibling PRs, full risk + review pane, optional codebase graph. No more inline-expand clutter.
Action labels (e.g. "Create Linear Ticket") and autonomy badges now render correctly in light mode. Intervention titles lead with the related PR or branch, e.g. "PR #48 · review scheduled".
Intervention rows render with the same agent badge, repository, branch / PR number, action label, policy name, and source link as Pull Request and push cards.
Agent log filters now appear inside the expanded agent log only — keeping the default surface uncluttered. The redundant compliance pill was removed; that count already lives in the top-bar chip.
Agent activity collapses to a metric strip. A new Decisions ribbon surfaces only what needs your action. The full activity history sits below — quiet on the happy path.
PR + push cards no longer expand by default for critical risk. Pinpoint summary stays visible inline; click to drill down when you want.
The card surface is cleaner. Hover the risk ring to see top risk factors with their scores; expand the card for the full breakdown.
Claude, Cursor, Copilot, Codex, Devin, CodeRabbit, Dependabot, Renovate, and bot badges now use darker shades in light mode for proper contrast.
Each compliance breach now identifies the exact PR, push, file, agent, or policy that triggered it, plus a tailored recommended fix. Breach packs land in PDF / JSON / CSV exports.
Merge / revert / incident outcomes now flow into your AI Act Article 15 evidence trail automatically — strengthening the calibration narrative auditors ask for.
The risk engine now uses an identical scoring briefing across `evaluate_change`, push assessments, and `pre_commit_check`. PR-specific context (title, ticket, CI status) is provided to the model as informational-only and explicitly excluded from scoring. A regression on this rule fails CI.
Risk Monitor, Batch Assessment, and the PR feedback prompt now display an `ANALYZING` indicator while the LLM is refining the score. Earlier versions could briefly surface the heuristic baseline before the LLM verdict arrived. This applies to PRs (gated on assessment phase) and pushes (gated on AI summary presence).
`evaluate_change`, `decide_action`, `pre_commit_check`, and `get_pr_code_review` are now capped at 30 calls per hour per token. Cheap reads remain unlimited. The 429 response carries a `retryAfter` field so CLI/agent clients can back off gracefully.
All MCP responses are now sanitized for ANSI/control characters at the API boundary, in addition to the existing CLI-side sanitizer. Terminals that are sent raw API output (custom MCP clients, curl pipelines) get safe text by default.
Code review findings are now dropped if the cited file path or line number is not actually present in the diff. Eliminates a class of LLM hallucination where a comment referenced lines outside what was sent for review.
CLI tokens now use a wildcard scope so newly-released MCP tools are usable without re-running `reentry init`. Fine-grained tokens with explicit allowlists continue to require an explicit scope update.
`evaluate_change`, `decide_action`, and `pre_commit_check` now always return an LLM-refined verdict. Synchronous calls block until the LLM completes (~25s), eliminating the case where a fast heuristic-only score reached you while the real LLM-refined score was still calculating in the background. Same diff, same answer, no surprises after the fact.
Four new commands close the CLI/dashboard gap: `reentry rules` shows your team's policies, `reentry review <pr>` renders the full AI code review (same content as the dashboard panel), `reentry fixes` prints agent-paste risk-reduction instructions (pipe to your coding agent), and `reentry log` lists recent assessments paginated. Read-only artifacts now live where developers work.
New `get_pr_code_review` MCP tool returns the complete structured review (AI summary, verify focus, inline findings by severity, suggestions, cross-file findings) for any PR your team monitors. Plus `list_recent_assessments` for paginated assessment history. Both available via Claude Code, Cursor, or the CLI.
`reentry agent add claude-code` and `reentry agent add cursor` auto-write the MCP server config so governance tools work in your IDE without manual setup. `reentry status` and `reentry explain` now auto-detect the repo from `git remote get-url origin`. JSON error envelopes for CI scripts. Tighter device-flow rate limiting and credentials file permission checks.
PR review cards now open instantly — the full AI analysis is pre-generated when GitHub fires the webhook, so the cache is warm before you click. Cards no longer show heuristic content while the AI is loading: only the AI summary, verify callout, code findings, and suggestions. Inline comments fall back as the card surface when the summary intentionally suppresses redundant key findings.
New Observe tab and `reentry observe` CLI command tail live agent sessions in real time — what an agent submitted, what context it sent, what risk verdict came back, and latency. Backed by per-team SSE with strict tenant isolation.
Guards that create Linear tickets on push events now execute correctly. Previously, a deferral loop prevented tickets from ever being created. Guards also now default to autonomous execution.
Push cards now display whether guard actions executed, are pending approval, or failed. Gives immediate visibility into guard effectiveness.
Fixed a silent bug where GitHub review IDs exceeding integer limits prevented the 'reviewed' status from being stored. Auto-posted reviews now correctly appear in the dashboard with their posted timestamp and link.
Code reviews are now pre-generated in the background as soon as the AI assessment completes. Opening a PR review is now instant instead of waiting 10-20 seconds for on-demand generation.
View AI findings directly overlaid on code diffs. Lines with issues glow by severity, click to expand details inline. Files sorted by criticality with auto-scroll to the first finding.
Watch any branch for risky pushes with a single click. Creates a governance guard automatically without manual policy configuration.
Auto-posted GitHub reviews now update in place when new commits are pushed, instead of creating duplicates. Score changes are posted as follow-up comments.
The blast-radius codebase graph in full assessment views is now collapsible, giving you more space for code review when needed.
Risk scores now stay consistent after AI refinement completes. Previously, scores could temporarily revert to preliminary values during periodic refreshes.
Enable automatic posting of severity-tagged, line-by-line code reviews to GitHub PRs when AI assessment completes. Toggle in team settings.
GitHub reviews now include severity badges on inline comments, risk dimension tables, file walkthroughs, and action checklists.
Both PR and push full assessment views now include the blast-radius codebase graph as a side panel for visual impact analysis.
Risk signals flagged by pattern matching are now automatically corrected when AI analysis determines they are false positives. Corrected signals appear dimmed with an explanation of why they were downgraded.
Push assessments now include a Full Assessment modal with risk drivers, AI findings, inline code notes, and review guidance — matching the PR assessment experience.
GitHub connection no longer requests broad repository write access. You now choose exactly which repositories to grant access to during a dedicated GitHub App installation step.
Connect GitLab repositories for merge request governance, risk assessment, and policy enforcement. Includes webhook monitoring, MR actions, and push guardian support.
Resolved a browser security policy issue that prevented the connected status indicator from appearing after completing OAuth in a popup. The integration card now reliably shows as connected.
Audit Log is now a dedicated top-level tab alongside Live, Intelligence, and Guards — no more buried sidebar button. The viewer has been redesigned to match the rest of the operational surface.
Code push reviews now include a one-click Copy Risk-Reduction Instructions button that builds a ready-to-paste agent prompt from AI findings and suggestions — the same feature previously available only on PRs.
A new get_risk_reduction_instructions MCP tool lets AI coding agents fetch structured fix instructions for any push or PR directly from the re-entry.ai API — enabling a fully automated assess-then-fix loop.
Apply AI-generated fixes directly from inline review findings. The fix agent creates a branch and PR for your review — no changes are merged automatically. Available on Solo, Pro, and Enterprise plans.
Code reviews now include the files that import your changed code, giving the AI reviewer call-site awareness to catch interface breakage, incorrect usage, and ripple-effect bugs that diff-only review misses.
The dependency graph now auto-zooms to search matches, shows all dependency edges at once, links directly to GitHub files, and refreshes automatically in the background. Monorepo packages are grouped into their own visual clusters.
Selecting a file in the codebase graph now lets you create a governance policy scoped to that specific file — no context switching required.
Intelligence is now a first-class tab alongside the live stream. Codebase scans run automatically every two weeks — no manual triggers needed. The first scan fires immediately when you enable intelligence.
Vulnerability detection now checks whether your installed package version is actually within the vulnerable range. Patched packages no longer trigger false alerts.
Blast radius signals (how many files depend on changed packages) were previously computed but not included in the final risk score. They are now a fully active scoring factor.
Precision, Recall, and False Positive Rate metrics now show contextual hover explanations. Incident and revert badges on risky files and hotspot directories also include tooltips.
Risk drivers now collapse by default showing only a severity dot and title — expand to see details. Impact radius now shows indexed file count rather than implying full codebase coverage.
Risk drivers in the code review panel can now be dismissed with a single click. Suppressed signals are excluded from the score immediately and remain suppressed even after AI refinement completes.
PR cards and the full assessment panel now display a live indicator while the AI is refining the preliminary score. The indicator clears automatically when analysis is complete.
The AI risk engine now retrieves similar past assessments, team correction patterns, and related incidents when scoring a pull request. Accuracy improves automatically as your team uses the platform.
Re-entry.ai supports flexible model backends — use our default, connect your own API key, or run a self-hosted model. Institutional learning carries over regardless of which model you choose.
View your entire codebase as an interactive graph. Files are color-coded by risk level, sized by importance, and organized by directory structure. Search, zoom, and click any file to see its risk history.
When risk drivers are identified, click 'Fix' to dispatch an AI agent that generates targeted code changes, creates a branch, and opens a pull request — all automatically.
Risk scores now appear within 500ms. The AI refines the score in the background and updates it automatically. No more waiting 10+ seconds for results.
The risk engine now tracks its own accuracy and automatically adjusts scoring weights based on real-world outcomes. Teams that use the system longer get more accurate results.
Disagree with a risk score? Override it with a documented justification. Active merge blocks are automatically removed. Overrides feed back into the calibration system.
See the full context behind every risk score: file incident history, author expertise, team baseline comparison, prediction accuracy metrics, and codebase learning progress.
Risk scores now account for how many files depend on the changed code and whether test files are included. High-impact files with many importers are flagged appropriately.
Missing Jira ticket links no longer inflate risk scores by 25+ points. Process compliance is now a minor signal, not a dominant risk factor.
Risk assessments now use AI to read and understand your actual code changes, dramatically reducing false positives. The AI considers intent, context, and team patterns — not just file names and line counts.
Deploy re-entry.ai with a self-hosted AI model. Your code never leaves your infrastructure. Supports Ollama and any OpenAI-compatible inference server.
Mark risk assessments as accurate or false positive. Your feedback helps calibrate the engine to your team's risk tolerance.
Settings pages now match the operational design language used across the dashboard. Cleaner cards, consistent spacing, and inline disconnect confirmations replace native browser dialogs.
Bitbucket OAuth now uses explicit scopes, proactive token refresh, and mandatory webhook verification for stronger security.
Risk intelligence now includes team health baselines, stale PR detection, and activity concentration warnings. Run active scans to check branch protection and unassessed PRs. Audit dependencies for known vulnerabilities across all monitored repositories.
The changelog now includes a search bar to quickly find specific changes by title or description.
Risk intelligence reports now include direct push risk assessments alongside PR assessments, giving a more complete picture of your team's risk profile.
The score history toggle on PR cards now shows the trend direction with color-coded icons — green for improving, red for worsening. Full assessment view shows the chart directly.
Risk assessments are no longer stored when the score hasn't changed since the last assessment, reducing noise and improving the accuracy of score history charts.
Generate a comprehensive risk intelligence report for your team showing prediction accuracy, risk distribution, coverage stats, risky files, hotspot directories, risk trends, and intervention summary. Includes GDPR-compliant anonymous mode.
PR assessments now show which changed files have a history of incidents or reverts. A visual risk map highlights the most dangerous files at a glance.
Risk assessments now learn which files have historically caused incidents or reverts. Files with a troubled history receive higher risk scores when modified.
Data files like changelog entries and configuration constants no longer trigger code injection warnings. SQL pattern matching now requires actual query structure.
Documentation and configuration files (.md, .json, .yaml) no longer trigger false code injection warnings. Lock-file-only PRs are now scored as low risk.
Risk assessments now detect Linear tickets in PR titles, descriptions, and branch names. Previously only Jira tickets were recognized.
Disconnecting GitHub now immediately clears all cached PRs, webhooks, and risk assessments. No more stale data after reconnecting.
Risk dimension labels now show clear levels (Low/Medium/High/Critical) instead of misleading percentages. Summaries explain why a score is high in plain English.
Webhooks from repositories you haven't selected for monitoring are now ignored entirely. Only your configured repos are analyzed.
Connect your Bitbucket Cloud workspace to enable PR risk monitoring, automated governance, and code review analysis. Full parity with GitHub integration.
Password reset emails now send reliably. Improved error handling ensures failed email delivery is caught and reported instead of silently failing.
Click the score change badge on any PR to see a visual timeline of how the risk score evolved across assessments.
Plan cards now use colored feature icons, cleaner layout, and consistent styling. Billing toggle spacing fixed.
Partner logos now load with a shimmer animation and scroll smoothly. Header links show active page state with a subtle background highlight.
Navigate signals with J/K, expand with Enter, focus the command bar with /, toggle chat with C. Press ? to see all shortcuts.
Risk summary now stays visible while chatting. Filters collapse behind a single button to reduce clutter. All commands route through AI for smarter intent handling.
Redesigned risk scoring explainer with animated L/I/D rings, updated integration showcase with horizontal scroll, new branded loader, and visual consistency improvements across all sections.
You can now reset your password via email. Click 'Forgot password?' on the login screen to receive a secure reset link.
Risk drivers now show clean severity indicators, human-readable descriptions, and collapsible file references instead of raw technical data.
Risk engine now detects when PRs target the wrong branch — like feature branches going directly to main instead of develop. Policies can auto-assign reviewers and request review on high-risk PRs.
The assign owner action now correctly resolves GitHub usernames and also requests a code review, not just assignment.
Policy names, descriptions, and GitHub label text can now be edited directly in the policy preview panel. Previously these fields were read-only.
Policy editor panels now scroll correctly so all fields are visible. Delete confirmations appear inline next to the item instead of at the top of the screen.
Accessing the dashboard or settings while signed out now redirects to the home page instead of a confusing signed-out confirmation screen.
Follow product updates in one place. Filter by category and scope to find the changes that matter to you.
Monitor real-time health of all Re-entry.ai services including API, integrations, and background workers with live status indicators.
Streamlined pricing to Free, Pro, and Enterprise tiers. Existing plans are automatically mapped to the new structure.
New step-by-step onboarding flow that connects your GitHub organization, selects repositories, and configures starter policies.
Risk scores now show compact concentric rings for each dimension instead of bars. Hover to see the weighting breakdown.
Long signal lists now collapse to the top 4 by severity with a toggle to expand. Duplicate signals are grouped with a count.
New pricing page, help center, about page, contact form, terms of service, blog, and FAQ — all publicly accessible.
Manage your password, notification preferences, and subscription billing from a unified settings page.
Subscribe, upgrade, downgrade, and manage your plan directly in the app. Powered by Paddle with automatic invoice generation.
Enterprise teams can now sign in with SAML/SSO via their identity provider. Configured through the admin portal.
Create, track, resolve, and write postmortems for incidents. Incidents link to related pull requests and trigger policy actions.
Risk assessments now factor in historical incidents, file sensitivity, and linked ticket context for more accurate scoring.
See risk scores, contributing factors, and recommended actions for each pull request in a single consolidated view.
Resolved issues where creating a second policy could show stale data, and editing policies now correctly loads all conditions and actions.
Policy-triggered actions like posting GitHub comments and applying labels now execute correctly when conditions are met.
Selected repository filter now persists when navigating between views instead of resetting to all repos.