Same pull request.
Two ways to spend
your afternoon.
Moved, extracted, renamed, modified — Mantish labels every change across files, then folds away the kinds you were never going to read.
One account, one rule set, every surface — and every diff is analysed on your own machine.
review, annotated
A pull request arrives as one pile of changed lines. Most of it is moves, renames and formatting you already agreed to — and it costs the same attention as the two changes that matter.
The panel to the left is the real review UI, not a screenshot — the same build the desktop app and the VS Code and JetBrains plugins host, running the actual classifier over a sample pull request.
Not everything that changed matters.
Every change gets an intent label. The noisy ones fold away until you ask for them. Click a rule to see what it catches.
usually muted
rarely muted
your call
Rename
Symbol renamed without logic changes
The built-in classifiers cover the usual noise. You can also write your own — in C# with full Roslyn AST access, or by describing in plain language what noise means to you.every capability, in detail
Four surfaces, one engine.
Pick the one that matches how you review. What differs is the shell — the classifier and the rules are the same in all four.
The same classifier ships in every surface — no lite build. Analysis runs locally in all of them: your code does not go to the host, and it does not come to us.
One account covers every surface. It carries the licence, not your code.what you can download today
Built for how you actually review.
Detects moved, extracted, renamed, and modified code — not just added and deleted lines.
See what changed at the class and method level. Expand any member to see the full diff inline.
Changed files as an interactive graph — which files reference each other, and where the clusters are.
Cluster diffs by category — every rename, every formatting tweak, every logic edit in its own bucket.
The same edit repeated across many files becomes one group — read it once, clear all of them.
Automatically dims formatting, imports, and whitespace changes so you can focus on logic.
The opposite of auto-mute — pin the change types you always want to see, so the risky edits stand out from the rest.
Browse changed files as a folder tree, or grouped by tag — tests, generated code, docs, infrastructure.
Follow the code you own — pull requests touching it reach you even when nobody added you as a reviewer.
Keep lockfiles and generated files out of counters, grouping and mute, so opening a pull request does not spend its time on them.
Rules live either in your own configuration or in the repository, where the whole team gets them from git.
C#, TypeScript, JavaScript, Go, Python, Java, C/C++, CSS and JSON are parsed structurally. Everything else still gets a line diff.
Native support for GitHub, Azure DevOps, GitLab, and Bitbucket — switch hosts without switching tools.
Review your own working tree the way you review a pull request — then hand the comments to a coding agent as a prompt and a checklist it ticks off.
A desktop app and a terminal client today; the VS Code and JetBrains plugins are built and waiting on their marketplaces.
Review threads are split automatically — a teammate's question in one list, the automated reviewers in another.
Every colour in the interface is editable, and a theme is a file — copy one, tune it, hand it to the rest of the team.
Protanopia, deuteranopia, and tritanopia palettes — diff colours that don't blur into each other.
Every name you wrote is replaced before anything reaches a model — keep proprietary code private.
Classification is deterministic — static analysis over the syntax tree, not a model guessing. AI only runs on rules you write yourself, against a model you point it at: we never run one over your code.each one, in detail
Your PR is bigger than your change.
Mantish is launching in 2026. Join the waitlist to get early access.
One email when it ships. No drip campaign.