Get ThruLine

Questions, answered plainly.

Six questions, a few sentences each. Every answer links to the page that shows it, so you can check the answer instead of taking it on trust.

Does ThruLine use an LLM to decide whether my code is correct?

No. An LLM — a large language model, the kind of AI that writes text — never decides whether a claim is proven. The verdict path is deterministic: the same record always gives the same answer.

The record is what your agent said and what actually ran. Every verdict rests on recorded runs of your own tests, which you can re-run yourself. After the verdict, a model on your machine (through Ollama) can narrate a finding in plain English. That is off by default, and it refuses any model that is not on this machine.

What runs where, on the Security page

What leaves my machine?

By default, nothing. ThruLine is local-first: it runs on your computer, and the record stays on your disk. Two things can leave, and only when you start them: a licence request when you fetch one, and, if you run thruline reexec, Docker downloading the container image it needs.

What Leaves this machine?
The record, runs and receipts Never. A receipt is a signed record of what ran.
Usage data (telemetry) Never. There is no telemetry.
Check it again (a re-run) Never. It installs nothing and uses no network.
A licence request Only when you fetch one yourself.
The optional narrator Never. It talks only to a model on this machine.
Everything that can and cannot leave, on the Security page

How is this different from CodeRabbit, Greptile or Cursor Bugbot?

They review the diff — the lines that changed — and some also run the code in a sandbox, an isolated copy of their own. ThruLine reviews the agent’s claim against the record of what actually ran, on the version — the exact commit — you are approving.

Tool What it does, from its own pages Source, fetched 25 Sep 2026
CodeRabbit Reviews pull requests — proposed changes to shared code — plus Change Stack, a review workspace. coderabbit.ai · pricing
Greptile Graph-indexed review, built on a map of your code; TREX runs the pull request in a sandbox. greptile.com · changelog, trex
Cursor Bugbot Cursor’s pull-request reviewer, with learned rules and Autofix, which proposes fixes. cursor.com · docs/bugbot
The full comparison, with sources, on the Product page

What does “Not proven” mean — is my code broken?

No. It means nothing in the record establishes the claim on this version. The claim may be entirely true.

It describes the evidence, not your code. ThruLine shows what it looked at and where the proof stopped, so you know what to run next. Only a different mark says a run failed:

thruline ui · marks · two kinds of Not proven
Not proven · version e71c5b0 Nothing in the record shows it.
Not proven · a run on e71c5b0 failed The record shows the opposite.
See one on the demo record: a file changed after the run

Can I use it on a team?

Pro covers one developer. Write to us if you need licences for multiple developers.

Pro, for one developer, is $15 a month or $150 a year. A licence unlocks one thing: approving a version. Recording, every read and every re-check never need one.

See Pro pricing

Is the source code public?

Yes. The source is on GitHub under FSL-1.1-MIT: you can use, modify and share it for anything except a competing commercial product, and each release becomes MIT two years after it ships. Pro is still what unlocks approving a version.

The source on GitHub

Open to beta users · Built by one engineer

Question not answered here? Ask the person who built it, or try ThruLine on your own change.

Every email is answered by the person who built ThruLine, within one business day. (Not a bot, not a queue.)

Review my change

or email support@thrulinehq.com