Mesrai is a multi-agent AI code review platform for product engineering teams. It reviews every pull request across GitHub, GitLab, Bitbucket, and Azure Repos using specialised AI agents for bugs, security, performance, architecture, and your team's own rules — usually in 2–3 minutes.

How does AI code review work?

Mesrai listens for new pull requests, parses the changed code into an AST, runs it through multiple specialised AI agents in parallel (bug detection, security, performance, architecture, custom team rules), then posts the findings as inline comments on the PR — ranked by severity. The full review usually lands in 2–3 minutes.

Yes for individual developers, as a launch offer: use the full review engine free on personal and open-source projects with your own AI key. Teams get a 14-day Free Trial that unlocks every Pro feature, no credit card required; after that, Pro BYOK is ₹499 / $6 per developer per month, or Enterprise on a custom contract.

Which git providers does Mesrai support?

GitHub, GitLab, Bitbucket, and Azure Repos — all four major providers. Setup takes about 2 minutes via the official OAuth app or a Personal Access Token. Self-hosted GitHub Enterprise Server is supported on Enterprise plans.

Will Mesrai store or train on my code?

Mesrai never trains models on customer code, and it is built for zero retention. Note: zero retention is rolling out — until it is enabled for your workspace, Mesrai keeps the pull request diffs and finding snippets needed for review history, and deletes them on request. With BYOK (Bring Your Own LLM Key), the AI processing runs on your own provider account.

What programming languages does Mesrai support?

JavaScript, TypeScript, Python, Java, Go, Rust, C++, C#, Ruby, PHP, Kotlin, Swift, Scala, and most modern languages. Architecture-aware review (via the mesrai-graph AST engine) is deepest on JavaScript/TypeScript, Python, Go, Java, and Rust.

Does Mesrai give engineering metrics like DORA?

Yes. Pulse — Mesrai's built-in analytics surface — tracks the four DORA metrics (deployment frequency, lead time for changes, change failure rate, mean time to restore) alongside PR cycle time, review depth, AI-finding adoption rate, and Mesrai-Rules coverage per team. Included on every paid plan; no separate dashboard subscription.

Multi-agent v2 — monorepo & multi-repo awareMulti-agent v2 · monorepo aware

AI code review that understands your architecture.

A multi-agent reviewer that maps your whole codebase as one graph — across monorepo packages and multiple repos, not just the diff. It traces the blast radius of every PR and turns it into DORA-grade metrics.

Start free See how it works

25,000+PRs reviewed·98%signal precision

● live review · pr #482

5 agents · whole-repo graph

[ The review bottleneck ]

Agents write code in seconds. Someone still has to review it.

Teams now open more pull requests than people can read. Mesrai reviews every one with full codebase context, so what reaches main is checked, explained and ready to merge.

Mesrai review

Pull requests from people and agents
Every change, full context
Verified · ready to merge

[ Whole-codebase graph ]

See the blast radius before you merge.

Mesrai maps every package and repo as one graph. A change in apps/web shows exactly what it touches across auth and payments — scored and flagged on the pull request.

MONOREPO · DEPENDENCY GRAPHPR #482 · feat: split checkout totalsIndex 1/4

indexing 6 packages · 156 files

changed reached finding not affectedfindings posted to PR #482

[ One review surface ]

Every capability, in its own view.

Code review, engineering metrics, the CLI and IDE sync — each a first-class surface your whole team works from.

[ Code review ]

Every PR, ranked by severity.

Mesrai reviews every pull request the moment it opens — explaining the reasoning and proposing a one-click fix, not just a vague comment.

Severity-ranked findings, each with rationale
One-click suggested fixes on the diff
OWASP / CWE security + performance checks

mesrai · reviewReview1/3

feat: add user authentication flowOpen#247 · 6 files

Mesraibotapi/users.ts:42 High · Security

XSS via URL parameter. searchParams is rendered into JSX without escaping.

Suggested change

−const msg = searchParams.get("error");

+const msg = escapeHtml(searchParams.get("error"));

CWE-79 · confidence 92% Commit suggestion

maya committed the suggestion3f2a9c1

Mesraichecked now

Verified. The value is escaped before render and the new test covers the XSS path.

agents · 5/5

Security
Performance
Architecture
Bug detection
Team rules

[ Pulse ]

DORA metrics, measured for you.

Every review feeds Pulse — your engineering-health dashboard. Deploy frequency, cycle time and bug ratio, scored against elite benchmarks.

Deploy frequency, PR cycle time, bug ratio
Scored against DORA elite benchmarks
Per-team and per-repo trends

mesrai · pulseDeploys1/4

Deploy frequency

0/week

Elite

PR cycle time

0h 00m

Elite

Bug ratio

0.0%

Needs focus

PR size (p75)

0 loc

Elite

Deploy frequencylast 12 weeks71%

5/wk · 12 wks agonow · 12/wk

suggestions applied92%last 30 days

[ CLI & CI ]

Review locally, before you push.

Run the exact same review from your terminal — as a pre-push hook or a CI step — and have the findings posted straight to the pull request.

Pre-push hook or CI step
Same rules as the cloud reviewer
Findings posted back to the PR

~/app · mesrai-cliRun1/3

~/app · mesrai-cli v0.4.25

❯mesrai review --branch feat/auth▋

resolving config · building graph · 847 nodes

5 agents · security · performance · architecture · bugs · rules

PASSarch/layer-boundary

WARNperf/n-plus-oneservices/orders.ts:88

FAILsecurity/xssapi/users.ts:42

posted 3 findings to PR #2471 fail1 warn1 pass

ci · passedposted 3 findings→ PR #247

[ IDE sync ]

One ruleset, everywhere.

Import the rule files you already keep for Cursor, Copilot, Claude and Windsurf. Your IDE, the cloud reviewer and the CLI all enforce the same policy.

Import Cursor / Copilot / Claude / Windsurf rules
No new config to maintain
One policy across IDE · cloud · CLI

mesrai · ide syncImport1/3

Cursor

12 rules

Copilot

9 rules

Claude

14 rules

Windsurf

7 rules

Mesrai Rules

42 imported · 4 duplicates merged

—rules

IDE assistantsIDE

Cloud reviewerCloud

Mesrai CLICLI

ruleControllers must not query the database directlyCLAUDE.md

synced4 rule files importedcursor · copilot · claude · windsurf

Works with the tools you already use

GitHubGitLabBitbucketAzure ReposVS CodeCursor

Trusted by fast-moving engineering teams

0K+PRS REVIEWED

0+TEAMS ONBOARDED

0%SIGNAL PRECISION

0–3 minREVIEW TIME

[ The problem ]

Code reviews don't scale with your team.

Senior engineers spend hours on PR feedback that catches the same bugs week after week. Reviews ship late, context gets lost, and security regressions slip into main.

beforeJunior PR sits for 2 days
Senior is in a meeting · waiting on a single approver
beforeSame bug class shipped
Async race in payment retry — caught in prod, not review
afterMesrai posts an inline review in 2–3 minutes
AGENT trace · 5 rules fired · 1 critical · 2 suggestions

[ How it works ]

From connected repo to merged with confidence.

Four steps, no config marathon. Connect once and Mesrai reviews every pull request automatically — inline, ranked by severity, with a recommended fix.

Connect repo

OAuth or PAT — no code changes. Set up in about 2 minutes.

GitHubGitLabBitbucketAzure Repos

Push a PR

Mesrai listens for the webhook and starts reviewing immediately.

Get inline feedback

Comments on the diff with severity, rationale, and recommended fix.

Merge with confidence

Auto-approve when clean · request changes when risk crosses threshold.

[ What Mesrai catches ]

Everything a senior reviewer would flag — automatically.

Specialised agents run in parallel on every PR — architecture, security, performance, your custom rules — then Mesrai turns the results into live metrics, shareable skills and memory your whole team builds on.

agent_graph

Architecture-aware review

Multi-agent analysis builds an AST graph of your whole repo — catching cross-file impact, layer violations and hidden coupling that linters and humans miss.

owasp · cwe

Security on every PR

SQL injectionXSSHardcoded secretsVulnerable depsIAM mis-scopes

p95

Performance regressions

N+1 queries, missed indexes, render thrash — flagged before prod.

pulse

Elite▲ 24%

DORA metrics, live

english · yaml

Rules in plain English

“All exported functions must declare an explicit return type.”

applied on 1,284 PRs · 4 repos

$0 markup

Bring your own key

OpenAI · Anthropic · Gemini

memory

Learns your codebase

Never re-suggests the same fix. Memory you control.

[ Why Mesrai ]

Not a linter. Not another bot.

Linters see one file. Most AI reviewers see only the diff. Mesrai reads your whole codebase as a graph, so it catches what the others structurally can't.

How Mesrai compares with linters and other AI reviewers
Capability	Linters & SAST	Other AI reviewers	RecommendedMesrai
What it reads	One file	The diff	Whole-codebase graph
Cross-file blast radius	Not traced	Limited	Traced across packages
Security · OWASP, CWE, secrets	Rule patterns	Varies	Every PR, ranked
Performance regressions	No	Some	N+1, indexes, p95
Your rules in plain English	Config files	Some tools	English or YAML
DORA & engineering metrics	No	No	Pulse dashboard
Same review in IDE + CLI	CLI only	Varies	IDE, CLI and CI
Bring your own AI key	Not applicable	Enterprise tiers	Every plan, $0 markup

Based on typical tools in each category. See tool-by-tool comparisons

[ CLI & IDE sync ]

Same rules for every reviewer.

Run the same review locally as a pre-push hook or CI step — and import the rule files you already keep for Cursor, Copilot, Claude, and Windsurf. One ruleset, everywhere.

// cli + ide_rules

Review locally. Sync your IDE rules.

Run Mesrai from the terminal as a pre-push hook or CI step — and import rule files you already keep for Cursor, Copilot, Claude, Windsurf, and more. One ruleset across every reviewer.

Explore CLI IDE rule sync

~ ❯ mesrai reviewv0.4.25

[12:01:01] PASS arch/layer-boundary
[12:01:02] INFO perf/n-plus-one services/orders.ts
[12:01:03] WARN security/sql-injection api/users.ts:42
[12:01:04] DONE 9 pass · 3 warn

[ Security & privacy ]

Your code stays yours.

Reviews run on your own AI key, in a sandbox that is thrown away when the review ends. You control the data path and the cost.

Where your code goes Your boundary

Your pull requestDiff and graph context
Ephemeral sandboxCreated per review, then discarded
Your AI providerYour key, your contract
Review on the PRInline comments, nothing kept

Never trains a modelNo third-party sharingZero retention · rolling out

Every plan

Encrypted in transitTLS on every connection
Encrypted AI keysStored encrypted, never shown again
Delete on requestReview history removed on request

EnterpriseEnterprise

Self-hostedIn your VPC or on-premise
SAML SSOSign in through your identity provider
Role-based accessDecide who can view and change what
Audit logEvery action recorded

Frequently asked

7 questions

What is Mesrai?
+
Mesrai is a multi-agent AI code review platform for product engineering teams. It reviews every pull request across GitHub, GitLab, Bitbucket, and Azure Repos using specialised AI agents for bugs, security, performance, architecture, and your team's own rules — usually in 2–3 minutes.
How does AI code review work?
+
Mesrai listens for new pull requests, parses the changed code into an AST, runs it through multiple specialised AI agents in parallel (bug detection, security, performance, architecture, custom team rules), then posts the findings as inline comments on the PR — ranked by severity. The full review usually lands in 2–3 minutes.
Is Mesrai free?
+
Yes for individual developers, as a launch offer: use the full review engine free on personal and open-source projects with your own AI key. Teams get a 14-day Free Trial that unlocks every Pro feature, no credit card required; after that, Pro BYOK is ₹499 / $6 per developer per month, or Enterprise on a custom contract.
Which git providers does Mesrai support?
+
GitHub, GitLab, Bitbucket, and Azure Repos — all four major providers. Setup takes about 2 minutes via the official OAuth app or a Personal Access Token. Self-hosted GitHub Enterprise Server is supported on Enterprise plans.
Will Mesrai store or train on my code?
+
Mesrai never trains models on customer code, and it is built for zero retention. Note: zero retention is rolling out — until it is enabled for your workspace, Mesrai keeps the pull request diffs and finding snippets needed for review history, and deletes them on request. With BYOK (Bring Your Own LLM Key), the AI processing runs on your own provider account.
What programming languages does Mesrai support?
+
JavaScript, TypeScript, Python, Java, Go, Rust, C++, C#, Ruby, PHP, Kotlin, Swift, Scala, and most modern languages. Architecture-aware review (via the mesrai-graph AST engine) is deepest on JavaScript/TypeScript, Python, Go, Java, and Rust.
Does Mesrai give engineering metrics like DORA?
+
Yes. Pulse — Mesrai's built-in analytics surface — tracks the four DORA metrics (deployment frequency, lead time for changes, change failure rate, mean time to restore) alongside PR cycle time, review depth, AI-finding adoption rate, and Mesrai-Rules coverage per team. Included on every paid plan; no separate dashboard subscription.

Get started

Ship code your team trusts.

14-day Free Trial. No credit card. Set up in about 2 minutes — first PR reviewed today.

Start free Read docs

AI code review that understands your architecture.

Agents write code in seconds. Someone still has to review it.

See the blast radius before you merge.

Every capability, in its own view.

Every PR, ranked by severity.

DORA metrics, measured for you.

Review locally, before you push.

One ruleset, everywhere.

Code reviews don't scale with your team.

From connected repo to merged with confidence.

Connect repo

Push a PR

Get inline feedback

Merge with confidence

Everything a senior reviewer would flag — automatically.

Architecture-aware review

Security on every PR

Performance regressions

DORA metrics, live

Rules in plain English

Bring your own key

Learns your codebase

Not a linter. Not another bot.

Same rules for every reviewer.

Review locally. Sync your IDE rules.

Your code stays yours.

Frequently asked

What is Mesrai?

How does AI code review work?

Is Mesrai free?

Which git providers does Mesrai support?

Will Mesrai store or train on my code?

What programming languages does Mesrai support?

Does Mesrai give engineering metrics like DORA?

Ship code your team trusts.