Privacy.

plottypus reviews your Claude Code sessions. To do that, the CLI sends your session to our server in Mumbai, and parts of it go to Anthropic’s model for review. We delete review traces 30 days after a session’s last activity and findings and session summaries after 90 days, never sell your data, and don’t train AI models on it. Separately, unless you turn off Help improve plottypus, we keep a shared copy of your reviewed sessions and scrubbed code snapshots for up to 2 years to test and improve plottypus. Content-free product metrics are always on.

Effective 22 October 2026. Version 2026-10-22.

Who we are#

plottypus is operated by Lay Sheth, an individual and sole proprietor in India.

• Legal name: Lay Sheth (trading as plottypus)
• Address: Siddharth Nagar, Indore 452009, Madhya Pradesh, India
• Email, for everything: hello@plottypus.com
• Grievance officer: Lay Sheth, same email and address
• EU and UK representative: not appointed yet. Write to us directly.

Status, 22 October 2026: the free trial is open and no payment is taken; paid plans are not open yet. When they open, we will sell to buyers in India first, and outside India (including the EU and UK) only once Paddle is live.

We are the controller for your account, billing, analytics and our logs. For other people’s data inside your code, see Other people’s data.

What we collect and why#

We never get: card numbers or UPI ids (Razorpay collects them), or your Claude or Anthropic login. Claude Code keeps running on your own plan; the CLI only adds hooks.

Account, session and usage data and product metrics are needed to run the service; billing data is needed to pay. Anonymous visit counts, product news, ratings and Help improve plottypus are optional, and turning them off changes nothing.

The CLI and what it reads#

When you connect a machine, the CLI adds hooks to Claude Code there. Then:

• It sends each Claude Code event on that machine to our server, except in folders you exclude or while the machine is paused.
• The reviewer can ask your machine to read files, list folders and search, only inside the folders of the session being reviewed. It can run only these read-only git commands: diff, log, blame, show. It can’t run any other command. Each answer is capped at 200 KB or 2,000 lines.
• It sends findings and checks back into your Claude Code session.

Your secrets stay out. Before anything leaves your machine, the CLI removes API keys, tokens, passwords and private keys, and it never reads .env files, key files, SSH or cloud credential folders, or anything your .gitignore ignores. Everything is encrypted in transit and at rest. Review traces are deleted 30 days after a session’s last activity, and findings and session summaries after 90 days.

Unless you turn off Help improve plottypus, the CLI also uploads a scrubbed snapshot of the repository at the start and end of each reviewed session, and we keep that shared copy of those sessions for up to 2 years. See Helping improve plottypus.

Keep sensitive repositories out. Don’t connect a machine, or pause it or exclude the folder, when a repository holds data you may not share, such as health, financial or government ID data: plottypus pause, plottypus exclude <path>.

Stop at any time: plottypus pause, revoke the machine in the dashboard or with plottypus logout, or plottypus uninstall.

Session data#

Session data can include:

• your prompts and slash commands;
• every tool call Claude makes, with inputs and outputs, including shell commands and their output;
• the full content of a file before and after Claude edits it (up to 1 million characters), and diffs;
• git status and git diff output;
• prompts and reports of Claude’s subagents;
• the working folder path and the rules files Claude loaded;
• files and git history the reviewer reads on request.

We use it to review that session for you. We don’t use it for advertising, we don’t sell it, and we don’t train models on it. We can open a trace to fix a fault or answer your request. Unless you turn it off, we also keep your reviewed sessions and code to test and improve plottypus (below).

Helping improve plottypus#

On the connect page, the Help improve plottypus switch sits above the Connect button, already on, with one sentence on what it keeps. Turn it off there and nothing is kept. Later you can turn it off for your whole account in Settings → Privacy, for one project with Don’t share this project, or from a machine with plottypus share off. Turning it off changes nothing about your reviews.

What we keep, for each reviewed session while sharing is on for you and that project:

• the session record: your task, every event the CLI sent, Claude’s messages, what the reviewer was given, and its findings and messages;
• a snapshot of the repository at the start and at the end of the session;
• the commits made in that repository since your last shared session, with their messages and changes, scrubbed the same way. The edits people make after Claude show us what the reviewer missed. They can include work by other people on the same repository.

What a snapshot holds:

• the files git tracks, or, in a folder without git, every file that isn’t ignored;
• never files your .gitignore ignores, deny-listed files, or secret files such as .env files, keys, and SSH or cloud credentials;
• every file goes through the same secret removal as your sessions before it leaves your machine;
• no binary files, nothing over 1 MB per file, and at most 50 MB per snapshot (past that, only the files that changed).

Paused machines, excluded folders, sessions with review switched off and projects you don’t share are never kept.

Why: to test and improve the reviewer. We replay real sessions against new versions, check which findings were right, what should have been flagged and what shouldn’t, and fix faults that show up across sessions. We may publish results that show none of your code or prompts, such as how often the reviewer was right.

Model training: we don’t train AI models on it today. If that changes, the model will serve only plottypus, and we’ll update this policy and email you at least 14 days first.

Where: our private storage in AWS Mumbai, India, encrypted at rest and in transit, closed to the public. When we replay a session to test the reviewer, parts of it go to the review model the same way as in a live review (see Model processing by Anthropic).

Who can access it: only us. AWS stores it for us, and parts reach the review model during a test. We never sell it, never share it with anyone else, and never let it be used to train another company’s general AI model.

How long: 2 years from the session, then deleted automatically.

How to stop and delete:

• Turn it off in Settings → Privacy, per project, or with plottypus share off. We stop keeping new sessions at once (for that project only, if you turned it off there). What you already shared stays until you delete it or the 2 years are up.
• Delete what I’ve shared, in Settings → Privacy, deletes every shared session, snapshot and commit copy from our live storage. Older stored versions expire within 30 days.
• Deleting your account does the same.
• Revoking a machine or pausing doesn’t delete what you already shared. Use Delete what I’ve shared.

Legal basis: the switch is in front of you, on, before you connect, with what it keeps. Leaving it on is your choice, and we treat it as your consent (GDPR and India’s DPDP Act). Where the law doesn’t count a switch that was already on as consent, we rely on our legitimate interest in testing and improving the reviewer you use: the copy is scrubbed, private, kept at most 2 years and never sold, and you can object any time by turning the switch off. We always honour it. Turning it off doesn’t make earlier use unlawful.

Code that isn’t only yours: while sharing is on, you confirm you may share that code with us for this. If it’s your employer’s or a client’s code, turn sharing off for that project, or check with them first. If you own code that someone shared with us without permission, email us and we delete it. For the shared copy we decide how it’s used, so we are the controller for it, including any other people’s data inside it.

Ratings and finding outcomes#

• On the dashboard you can mark any finding Helpful or Not a problem, add a note, and answer Was this review useful? at the end of a session.
• We also record what happened to each finding: whether Claude changed the flagged lines, whether the re-check came back clean, whether Claude argued against it, whether it was still there when the session ended, and how long a fix took.
• We use these to measure and improve the reviewer. They are kept with the finding, counted in product metrics without any text or note, and, for a shared session, kept with the shared copy.

Other people’s data in your code#

Your code, fixtures, logs and git history can hold other people’s personal data, such as your customers, coworkers or open-source contributors (git log, blame and show return commit author names and emails). For that data we act as your processor under our data processing addendum, which is part of the terms for every customer. The exception is the copy you share through Help improve plottypus, which is on unless you turn it off; for that copy we are the controller.

Model processing by Anthropic#

To review a session, our server sends parts of it to an AI model made by Anthropic: event summaries, prompts, diffs, file excerpts, tool results and earlier findings. Any session data above can reach the model. Our server runs it with no tools of its own and no saved conversation on our side.

Until our first paying customer, reviews run on Anthropic’s consumer Claude service, under our own Claude account, with model training turned off in that account’s privacy settings. That means:

• Anthropic’s consumer terms and privacy policy apply to this data, not its commercial terms. We have no data processing agreement with Anthropic for it, and no zero-retention agreement.
• With training off, Anthropic keeps consumer data for 30 days. Anthropic’s own privacy policy sets any exceptions.
• Anthropic processes it in the United States.

Before we take the first payment, we move reviews to Anthropic’s commercial API or to Amazon Bedrock. We’ll update this page and email you the day it happens.

Where your data is stored and sent#

• Our server, storage and backups: AWS, Mumbai, India. Encrypted disk and encrypted backups. Shared sessions and snapshots are in private, encrypted storage in the same region.
• Accounts and sign-in: SurrealDB Cloud and Vercel (site functions in Mumbai).
• Dashboard: served by Cloudflare. Your session data reaches your browser straight from our API in Mumbai.
• Model: Anthropic, United States.
• Product metrics and anonymous visit counts: PostHog, United States.
• Email: Brevo, with Resend (US) as a fallback. Mail to us is routed by Cloudflare.
• CLI downloads and updates: GitHub releases (US) and dl.plottypus.com, served through Cloudflare.
• Payments: Razorpay in India. Paddle (UK) outside India, once live.
• Sign-in: GitHub and Google, US.

The full list is on the subprocessors page.

Transfers out of the EU and UK. India has no EU or UK adequacy decision. We rely on the EU standard contractual clauses and the UK Addendum in each vendor’s DPA, and on the EU-US Data Privacy Framework where a vendor is certified. Anthropic is the exception for now: under the consumer service there is no DPA between us and Anthropic. Ask us for a copy of the safeguards.

Transfers out of India. India’s DPDP Act allows transfers except to countries the government restricts. We know of none today.

Analytics, cookies and recordings#

Product metrics#

To run, measure and fix plottypus, the CLI and our server record product metrics. They are part of the service, so they’re always on; the way to stop them is to delete your account.

They never contain your code, your prompts, Claude’s text, file contents or file names. Folder paths and repository names are hashed with a secret key before we store them, and only file extensions are kept. Metrics are linked to your random plottypus user id, never your email or name.

What they cover:

• setup: install page seen, machine connected, first session, first review, first finding, first fix, active trial days, trial end, plan start;
• each session: how long it ran, git or not, a rough file count, the mix of file types, how many edits, files and lines changed, tool calls by type, test runs with pass and fail counts, how often Claude said it was done, Claude Code and CLI versions, OS and chip type, and your plottypus settings;
• each review: what started it, how long it took, the model, tokens, cost, and findings by level and kind;
• each finding: level, kind, file type, how and when it was delivered, and what happened to it;
• each hold at “done”: how long, why, whether it timed out, and whether you typed while Claude was held;
• each message to Claude: whether it arrived, whether Claude acknowledged it, and whether code changed after;
• reliability: hook errors, timeouts, restarts, network fallbacks, upload and update failures, versions in use;
• dashboard: changes to your controls (pause, mode, excluded folders, review of new sessions, sharing) and feedback clicks.

Where they go: our server in Mumbai keeps them and forwards them to PostHog (US), our processor, from the server. These events carry no IP address, and nothing is stored in your browser for them.

Legal basis: contract, to meter and run your service, and our legitimate interest in keeping plottypus working and making it better.

How long: on our server until you delete your account; in PostHog 12 months.

Anonymous visit counts#

No banner, because nothing is stored on your device. On the site and in the dashboard we count page views and button clicks anonymously, to see which parts work. Nothing is stored in or read from your browser for this: no cookie, no local storage.

• PostHog (US) gets the page, the button, and your device and browser type. It uses your IP address only to work out your country and region and to make a visitor hash that changes every day, and then drops the address. We never keep the IP address, city or finer location.
• These counts are not linked to your account, name or email. What you do in plottypus while signed in is counted by our server as product metrics, described above.
• The link that brought you here travels with the page view, not in a cookie.
• Legal basis: our legitimate interest in knowing which pages work. To opt out, open “Privacy choices” in the footer, or dashboard Settings → Privacy, and turn off “Count my visits anonymously”.
• PostHog keeps these events for 12 months.

No screen recording. We don’t record your screen on the site or in the dashboard.

Cookies we always set, because the service needs them: the sign-in session cookie, a 10-minute cookie that carries your sign-up agreement through GitHub or Google sign-in, and, only if you turn off visit counts, a pp_consent cookie (6 months) that remembers that choice. Once paid plans open, Razorpay’s payment window may set its own cookies.

No advertising cookies. We don’t sell or share your personal information.

How long we keep it#

Your rights#

Wherever you live, you can:

• Get a copy: Dashboard → Settings → Your data → Download gives your sessions, findings, usage, machines, plan and consents as one JSON file. For a trace still kept, email us.
• Correct: email us, and we fix it.
• Delete: Dashboard → Settings → Your data → Delete account. It revokes every machine, cancels your plan and its mandate, and deletes your account, machines, sessions, findings, usage, traces, product metrics, and any shared sessions and snapshots from our live systems. We delete your Brevo contact at once, and ask PostHog to delete you and your events within 30 days. Backups expire 30 days later. We keep billing records 8 years (tax law), consent records 3 years (proof), and a hashed email 2 years for the one-trial rule. To delete one computer’s sessions, use Settings → Machines → Delete sessions; to delete one session, email us.
• Withdraw consent or object: Help improve plottypus, turn it off in Settings → Privacy, per project, or with plottypus share off, and use Delete what I’ve shared to remove what you already shared. Product metrics have no switch, because they are how we run and fix the service; to stop them, delete your account. You can still object to them under GDPR: write to us. Analytics, any time in Privacy choices or Settings. Session data: disconnecting a machine (in the dashboard, plottypus logout or plottypus uninstall) stops new data; its sessions stay until you delete them under Settings → Machines. Deletion runs 7 days after you ask, and you can undo it until then. Pausing stops sending but keeps what was sent.
• Object, restrict, or take your data elsewhere (the download is machine-readable JSON).
• Nominate someone to act for you if you die or can’t act (India’s DPDP Act). Email us their name and contact.
• Complain: to us first. You can also complain to the Data Protection Board of India, your EU data protection authority, or the UK ICO.

Use the dashboard, or email hello@plottypus.com. We may ask you to confirm from your account email. We acknowledge within 48 hours and answer within one month, free. A complex request may take two more months under GDPR; we’ll tell you why within the first month.

The reviewer gives advice on your code. It makes no decision about you with legal or similar effect.

Children#

plottypus is for people 18 or older. You confirm this when you sign up. We don’t knowingly collect data from anyone under 18, and if we learn we have, we delete the account and its data.

Security#

What we actually do:

• Our server’s disk and backups are encrypted.
• All traffic to the site, dashboard and API uses HTTPS.
• Machine tokens and one-time codes are stored only as hashes on our server. On your machine the token is a file only your user can read, or the macOS Keychain if you choose.
• Trace files are readable only by the service account and by us, as described above.
• Shared sessions and snapshots sit in private storage that blocks all public access. The CLI uploads each snapshot through a short-lived link for that one file, which our server issues only while sharing is on for you and that project.
• The reviewer reads only inside the session’s folders, never deny-listed files, and runs only read-only git commands.
• Server logs remove tokens and known key formats.

No system is perfectly secure, and we don’t claim ours is. To report a problem, see the security page.

Breaches#

If a breach affects your personal data, we tell you without delay: what happened, the likely impact, what we did, what you should do, and who to contact. We tell the Data Protection Board of India without delay, with a full report within 72 hours, and the relevant EU or UK authority within 72 hours where GDPR requires it.

Changes#

The version is at the top. We keep every earlier version and send it on request. For a change that affects data you already gave us, or how we use it, we email you at least 14 days before it takes effect, and ask again where the law needs it.

Grievance officer and contact#

• Grievance officer: Lay Sheth
• Email: hello@plottypus.com
• Address: Siddharth Nagar, Indore 452009, Madhya Pradesh, India

We acknowledge every complaint within 48 hours and resolve it within one month. This policy is in English; ask us for a version in any language in the Eighth Schedule of the Constitution of India.