Code Review
Overview

Code Review

AI review on every pull request

A linter checks the file in front of it. A human reviewer, most of the time, does too — nobody has time to trace every call site a change might touch before approving. Entelligence reads your whole codebase, not just the diff, so it catches the things that only show up when you look at how a change fits everything around it.


🎥 Video demo — coming soon

Watch a live PR get reviewed, findings triaged, and a fix applied


Without full context vs. with it

❌ Diff-only review

Sees a function signature changed. Misses that three other callers now pass the wrong number of arguments, because those callers live in files that weren't touched.

✅ Entelligence

Traces the signature change to every caller across the repo, flags the ones that will now break, and points at the exact line in each.

That's the difference between checking a diff and understanding a change.

What shows up on your PR

  • A summary comment — a plain-English walkthrough of the change, key findings, and which files need the most attention.
  • Inline comments on the exact lines that matter, each tagged by severity — CRITICAL, MAJOR, or NIT — so you know what actually blocks a merge versus what's a nice-to-have.
  • Committable suggestions — many findings ship with a ready-to-apply code suggestion you can commit straight from the PR.
  • "Prompt to fix with AI" — a ready-made prompt attached to each finding that you can hand to Claude Code, Cursor, or any coding agent to fix it directly.
  • A Confidence Score (1–5) — from "Safe to Merge" down to "Blocking Issues" — computed from the severity and resolution state of every finding, so you can tell at a glance whether a PR is ready, without reading every comment.

Reviews post within moments of a PR opening — no waiting on a queue, no "runs nightly." Every subsequent push gets a fresh pass, and issues you've already resolved don't come back unless the underlying code actually regresses.

Priority levels

Set per repo from Codebases, this controls how much gets surfaced — so a repo full of experiments doesn't get the same bar as your payments service:

LevelWhat gets flagged
CriticalOnly the most severe, high-confidence issues — least noise, highest bar
MediumMedium-priority issues and above — a reasonable default for most teams
LowEverything, including minor/nitpick-level findings

Talking to the bot

@mention the bot in any PR comment, or just reply directly to one of its own comments — no mention needed there.

Say thisIt does this
@entelligence reviewRuns a fresh full review
@entelligence summaryRegenerates just the PR summary
@entelligence resolve_allResolves every open bot thread at once
@entelligence configPosts the bot's current settings for this repo
@entelligence link <PR>Links a related PR in another repo for cross-repo context
Anything elseAnswered directly, using the PR as context

It learns your team's standards, not generic ones

Two feedback loops, working together:

  • Thumbs down a finding that isn't actually a problem, and Entelligence factors that in — recurring false positives fade out instead of repeating every PR.
  • Upload a guideline doc (your own style guide, or a language best-practices doc), or have Entelligence generate one automatically by mining your repo's own review-comment history. Either way, review comments start reflecting your conventions, not a generic checklist.

Language coverage

Built-in, language-aware review guidelines cover Python, JavaScript/TypeScript (including JSX/TSX), Java, Go, C, C++, C#, Ruby, Rust, and Dart — on top of general cross-language checks that apply regardless of language.

Dashboard

TabWhat it shows
OverviewRecent PRs, review status, and confidence scores at a glance
Code QualityTrends in finding severity and volume over time
Review AnalyticsTurnaround time, comment acceptance rate, and bot activity
CodebasesPer-repo settings: priority level, guideline docs, and enable/disable

Two ways to run it

Automatic PR reviewCLI review
TriggerPR opened, updated, or marked ready for reviewYou run entelligence review
Where results appearInline comments + summary comment on the PRPrinted in your terminal
Best forEvery PR on a connected repo, automaticallyLocal pre-push checks, repos not yet connected, non-Git-hosted workflows

Both paths run the same underlying review engine — see Reviewing from the CLI for the terminal workflow.

Supported platforms

GitHub, GitLab, and Bitbucket — including self-hosted GitHub/GitLab instances.