Open source. Always yours.

Ship the code.Catch thebugs.

AI code review that understands your codebase.
Catch what matters, skip the noise, and merge with confidence. On infrastructure you control.

Free, Apache 2.0 Bring your own model
acme / platform
Harden the authentication flowOpen #248 · alex wants to merge 3 commits
src/auth/session.ts+12 −4
38export async function logout(userId) {
39− session.revoke(userId);
39+ await session.revoke(userId);
40}
Mirabotjust nowPotential bug

Wait for the session to be revoked.

Without await, the audit event can finish before the session is invalidated. The token may still be usable after logout.

Suggested change+ await session.revoke(userId);
Review complete8 files reviewed / 3 findings
The context behind the code.Dependencies, conventions, and the whole repo.
ILLUSTRATIVE REVIEW · REAL MIRA CAPABILITIES
Built for your workflow.
Never the other way around.
GitHub GitLab Forgejo
A second set of eyes. A whole lot of context.

Less noise.
More “good catch.”

Go beyond the diff. Mira connects your changes to the rest of your codebase, then leaves feedback you can act on.

01 / THE REVIEWMake it work your way
acme / platform
Harden the authentication flowOpen #248 · alex wants to merge 3 commits
src/auth/session.ts+12 −4
38export async function logout(userId) {
39− session.revoke(userId);
39+ await session.revoke(userId);
40}
Mirabotjust nowPotential bug

Wait for the session to be revoked.

Without await, the audit event can finish before the session is invalidated. The token may still be usable after logout.

Suggested change+ await session.revoke(userId);
Review complete8 files reviewed / 3 findings
SELECT A REVIEW TYPE TO EXPLORE AN EXAMPLE
02 / YOUR INFRASTRUCTURE. YOUR RULES.

Your code.
Your model.
Your call.

A great code reviewer shouldn’t mean giving up control. Run the full Mira engine on your own infrastructure, with the model that works for your team.

Explore deployment options
YOUR INFRASTRUCTURE
Your repositories Your pull requests
MiraIndex. Review. Learn.RUNNING
Your choice of modelLocal endpoint or your own provider key
Your review history Your data store

Keep everything local with a local model. With a cloud model, review context is sent to the provider you choose.

Open from the ground up

Apache 2.0. Read the source, change the rules, and make Mira your own.

Any model. No markup.

Use your own API key or local endpoint. Pay your model provider directly.

The whole product, self-hosted

The review engine, indexing, dashboard, and learning loop. All included.

03 / FROM PULL REQUEST TO PEACE OF MIND

Fits right in.
Gets smarter.

No new review ritual. Open a pull request, get useful feedback, and keep working where you already do.

01

Connect the context

Mira indexes files, symbols, and dependencies to understand how your code fits together.

02

Review what matters

Specialized review passes find issues, filter noise, and leave actionable comments on your pull request.

03

Learn with every merge

Resolved feedback and merged PRs help Mira adapt to your team’s conventions over time.

04 / KEEP YOUR MOMENTUM

Thoughtful reviews.
Without the wait.

Speed matters when you’re ready to merge. Explore Mira’s recorded results on the public Code Review Bench.

50pull requests
5codebases
3review metrics
Explore the full results
MIRA / RECORDED RESULT
77s

Median review time per pull request

Lower is better
50 pull requests · Offline benchmarkMethodology

Recorded results from the Mira repository, measured on the 50-PR offline set of the Martian Code Review Bench: 10 PRs each from Sentry, Grafana, Keycloak, Discourse, and Cal.com, judged by Claude Sonnet 4.6. Competitor figures come from the public dashboard and review times are approximate. This is a selected comparison of four reviewers, not the full leaderboard.

View results and methodology
05 / GET UP AND RUNNING

Your next reviewer.
Your deployment.

Choose where Mira runs. Bring your model and connect your repositories. The full product is yours from day one.

Create your mira.yaml and .env with your model and Git provider credentials first, then start the container.

TERMINAL
docker run -p 8000:8000 --env-file .env \
  -v "$(pwd)/mira.yaml:/app/mira.yaml" \
  ghcr.io/miracodeai/mira:latest --config /app/mira.yaml
A LITTLE MORE CONTEXT

Good questions.
Clear answers.

Everything else you might want to know before your first review.

Explore the docs ↗
Is the whole product really free?
Yes. Mira is open source under Apache 2.0. Self-host the review engine, codebase index, dashboard, vulnerability scanning, and learning loop. There is no license key or paid feature tier. You cover your infrastructure and any usage billed by your chosen model provider.
Does my code leave my infrastructure?
Mira stores your indexes, review history, and configuration on infrastructure you control. If you use a local model, review context stays local too. If you choose a cloud model, the context needed for a review is sent to that provider using your own key.
Which models can I use?
Use OpenRouter or an OpenAI-compatible endpoint, including local models served by Ollama or vLLM. Mira also supports direct AWS Bedrock connections. Choose the provider and model that fit your privacy, cost, and quality requirements.
Does Mira work with our Git provider?
Mira supports GitHub, GitLab, and Forgejo, including Codeberg. Connect your deployment using a GitHub App or the appropriate provider access token to start reviewing pull requests and merge requests.
How do I get started?
Deploy Mira with Docker, Railway, Fly.io, or Render. Configure your model and Git provider credentials, then connect your repositories. The quickstart walks through the setup, including creating a GitHub App for GitHub repositories.
BETTER REVIEWS. ON YOUR TERMS.

Make your next merge
a confident one.

Open source. Self-hosted. Ready for your codebase.