Every part of the board, and why it’s there.
Limn is opinionated about one thing: the fastest way to understand a team is to see it. Everything here is free and open-core — not a trial, not a taster. The core is the product.
Caught up in ten seconds.
One cross-repo stream of what actually happened — opens, merges, reviews, and the commits that addressed a thread — coalesced per author, bot-filterable, full markdown inline. Click through to the PR; Back returns you to the exact item you left.
“Some bots are helpful but their automatic comments add noise to my notifications/emails and there seems to be no way to configure that.”
08:58 · Coffee down, feed open: two merges overnight, one question addressed to you. That’s the whole catch-up.
No feedback left behind.
During sync, Limn classifies every review thread by cross-referencing replies and resolution against the commits that landed afterward. One state is a heuristic — and the product never pretends otherwise.
“In 37 (74%) PRs, the reviewer responded after one week but quickly merged the PR after reviewing it.”
Most late reviews aren’t hard reviews — they’re unseen ones. A thread state you can scan is the difference.
Resolved
Marked resolved on GitHub. Done — no action needed.
Likely addressed
A commit touched the thread’s file after the last comment. A heuristic — and the UI says so, because renames and unrelated edits can fool it.
Replied · unresolved
Someone replied, but it’s still open and no later commit touched the file. A live conversation with no fix yet.
Untouched
No reply, no follow-up commit. Feedback left and, so far, ignored.
Day 3 · The critical PR’s thread flips to untouched. Someone answers it today — because it was visible, not because anyone nagged.
Repos down the side. Time across the top. Everything in between.
Each repo holds a row per contributor; PR bars pack into lanes; events render as shaped markers that cluster as you zoom. A list tells you a PR exists — a timeline tells you it’s been open eleven days, reviewed once on day two, silent since. Duration and staleness live in the shape.
“…the authors had to grab the attention of the reviewer by mentioning their PRs in the comment of other PRs.”
Sprint day 9 · The migration PR’s bar is long and quiet. You see it, you ping once, it ships — no nagging economy required.
Standup, pre-assembled.
Pick a repo: a stats header, a thread-state bar showing how much feedback is resolved versus sitting, every open PR with its CI and approval standing, that repo’s own feed underneath. Re-scopes live with your filters.
09:58 · Two minutes before standup, “how’s the payments repo doing?” is already answered — before anyone shares a screen.
The loop stays closed.
Overview (CI checks with failing-job logs, reviewers vs approvers), threads grouped by file, inline diffs with unresolved threads pinned to their lines. And it’s not read-only: reply, resolve, @mention, request reviewers, approve, rebase, merge — real GitHub writes, gated on your real permissions.
“…you have your mind on something else and then you have to shift completely… by the time you come back to it your frame of mind is completely different…”
Mid-review · Read, reply, approve, merge. The tab count never moves; neither does your frame of mind.
Noise, measured.
Every bot comment graded by Limn’s own ML model — severity and category, no vendor self-assessment. Rolled up per bot: cost, noise mix, overlap, and the bot-only reviews no human ever handled. Grading runs in the hosted service today; the local install is on the roadmap.
“When we first launched this product, the biggest complaint by far was that the bot left too many comments.”
Renewal week · “The bot feels noisy” becomes keep, tune, or kill — decided on evidence, calmly.
Found, not remembered.
Cross-repo search, instantly — PRs, threads, comments, people — against the locally-synced index, results as you type. It’s the / key. Workspaces group repos by team: every view scopes to one, bots are judged per workspace, and the compare board tracks metrics across all of them.
“People couldn’t find things. It was simple as that. It took forever to just find the right service.”
Thursday · “Which repo was that thread in?” stops being a question anyone asks.
Fast enough to be a habit.
Everything syncs into a local database behind a deliberately lean read layer — the board never waits on GitHub to render, and adaptive sync keeps hot repos seconds-fresh without burning quota on cold ones. Filters compose instantly, and every view is a URL — shareable, bookmarkable, pasteable into a standup note.
Every day · A dashboard you wait for is a dashboard you stop opening. This one opens fast enough that you actually do.
Hands on the keys.
- /Jump to the repo search
- jkCycle through PRs
- iOpen the Activity console
- EscStep back out — tab, then selection
Pro reads the board. Pro+ acts on it.
Digests with teeth, thread validity, themes and reports, chat with your repos — and in Pro+, the Claude loop: review, fix, push, one app.