Quick answer
plottypus is open to everyone. It reviews your Claude Code session while it works, checks every “done” against what actually ran, and messages the agent with the proof when it does not hold. The free trial is 7 days or 10 review hours from your first connect, with no card. Start with one install line.
What plottypus does
Coding agents say done when the work does not hold. The tests never ran after the last edit, a command failed and scrolled past, or the change is not what you asked for. A pull request review sees only the finished diff, so it cannot tell.
plottypus follows the whole session: what you ask, each edit, each command and its output, and when the agent says done. The reviewer checks that claim against what actually ran. When it does not hold, plottypus tells the agent, with the output that shows it, while the agent still remembers why it wrote the code. The agent fixes it. Everything else lands in one review file. It leaves out style, naming and tidy-up notes, so each comment is one worth reading.
It never changes your code. Your agent makes every change, under the permissions you already gave it.
The trial
- 7 days or 10 review hours from the first time you connect a machine, whichever comes first.
- No card needed.
- Paid plans open later: one plan for 25 hours of review a month, priced in dollars and in rupees on the pricing page, with a hard cap so you are never charged more.
- If your trial ends before paid plans open, write to hello@plottypus.com and we’ll keep you going.
How to start
- Press Start free trial on plottypus.com and sign in.
- Paste one line into your terminal:
curl -fsSL https://plottypus.com/install | sh- Your browser opens to connect the machine. Then keep using Claude Code exactly as you do now.
It works on macOS, Linux and Windows (WSL). Pausing, excluding a folder, updating and uninstalling are each one command: they are all on the setup guide.
How we test it
We publish how we test plottypus, misses included, on /evals. One result from there: on two sets of real pull requests from the Martian Code Review Bench, with every tool’s comments scored by the same judge, plottypus ranked third of eight on F1. The page gives the date, the sets and the sample size for every number, and what we are still improving.
What we keep
We delete review traces after 30 days and findings and session summaries after 90 days. Separately, unless you turn off Help improve plottypus (when you connect, in Settings → Privacy, per project, or with plottypus share off), we keep a shared copy of your reviewed sessions and scrubbed code snapshots for up to 2 years to improve plottypus; the details are on the privacy page.
Tried it and something broke, or it missed a false done? Write to hello@plottypus.com. A person reads every message.