Limn — automated, local code review
“The line without shadow showeth all to good judgment; but the shadow without line showeth nothing.”
Code review, story alignment, docs alignment. Automated, on your machine.
Free. Local. No extra seat.
Quick demonpx limn-review
Gather. Check. Act.
The pull request, its story and your docs, in one place.
Your own AI agent reviews the change against all three, on your machine.
Reviews, fixes and posts run on their own. You merge.
Limn is an advanced harness for code review: your tools, your AI agent and one clear screen, working as one. It does the checking, so your team ships faster, for no extra cost.
Gather
Limn reads each tool where it already lives and keeps a copy on your disk.
Bring together the tools you already use: your code host, AI agent, tracker, internal documentation and review bots.
Check
Your AI agent reads each pull request against its story and your own docs, and says which one it broke.
RefundPolicy.allows() never reads order.paid_at.
amount is stored as a float. The handbook says whole pence, in an integer column.Your docs are a library you build in Limn from your own pages. Every finding cites the section.
See what it changes
It draws the parts a pull request touches. For a story, one drawing across all its pull requests.
Limn draws one for every review, made for that change. Each part cites the files behind it, so you can follow the key ideas straight to the code.
Time it saves you
The repetitive parts of review, done for you, so people can spend their attention on the change itself.
Pasting diffs into a chat window
Limn reviews each new push on its own, and can post the findings to the pull request for you.
Re-reading the ticket
Each change is checked against the story’s acceptance criteria, criterion by criterion.
Looking up the house rules
Findings cite the section of your own docs that the change goes against.
Digging through CI logs
A failing check is diagnosed from its log, and AI Fix drafts the change.
Re-reviewing after every push
Each new commit is reviewed once, with a note on which earlier threads it addressed.
Working out who is waiting on whom
Every open pull request is ranked by whose turn it is, across every repository.
Act
Limn reviews every push, posts the findings to GitHub and fixes what it can. Merging stays with you.
Faster reviews, fewer bottlenecks
On a nine-repository delivery team, Limn took over the reviews in October. Half of their pull requests now merge within an hour of working time, and twice as many land each week. It cost them nothing extra.
Measured by Limn's Chronology report: merged pull requests by people, working hours, September vs the week of 19 October 2026.
No extra seat
No per-review fee, no new bot to pay for. Limn runs on the AI agent you already have: Claude Code or Codex.
Want a ceiling? Give Limn a share of your weekly AI usage, 5% unless you change it, and automatic reviews stop when it's used. Set a budget
What dropping a review-bot seat saves
Worked out in your browser. Nothing is sent.
Free to use
| Up to 3 peopleFree | 4 or moreCommercial licence | |
|---|---|---|
| Every feature | Included | Included |
| Model usage | Your own AI agent | Your own AI agent |
| Runs on | Your machine | Your machine |
| Run it now | Get in touch |
Questions
Nothing from us. Reviews run on your own AI agent — Claude Code or Codex — and Limn shows what each run used.
No. Everything lives in SQLite on your disk. Your AI agent sees code only when a review runs.
Yes, once you switch on auto-posting for a workspace. It posts one review to the pull request.
Yes. Limn reads their comments and ranks them beside your own AI agent's.
Enterprise Cloud on github.com, yes. Self-hosted Enterprise Server, not yet.
