Updated Oct 31, 2026

Privacy

Cactus reads the sources a team connects. This page says what we hold, what we do with it, how long it stays and who else touches it.

What we collect

  • Your account. The name and email on the account you log in with.

  • Your plan, if you pay for one: which plan, its status and its dates, and the ids Polar uses for your team. Your card details go to Polar, never to us.

  • The sources an owner connects to a product: code repositories, a help site and Intercom conversations.

  • Messages people send to Cactus on Slack and on Discord, and the team's replies in a thread Cactus handed to them.

  • Conversations people have with Cactus, including the ones a visitor starts on a published Help Center without an account.

  • Usage events and errors: which pages get opened, which questions get asked, and whether an answer worked.

  • Where you came from, if a link brought you: for your first visit from another site, a tagged link or an ad, and for your latest from a tagged link or an ad, the address you landed on (with its campaign tags and any ad's click id), the site that sent you and the time. It is kept with your waitlist entry, and with your account and team if you start one. Your waitlist entry also keeps the random id our analytics gave your browser, so the visits before you joined can be matched to it.

Some of what reaches us is never used. Slack sends an event for every message in a channel Cactus has joined. On Slack and on Discord we keep only the messages sent to Cactus and the team's replies in a thread it handed to them, and the rest are dropped without being stored. Cactus cannot read private Slack channels, or direct messages other than the ones sent to it.

How we use it

We use it to answer questions and write pages for that product, your plan to bill you for it, and where you came from to count which of our posts and ads bring people to Cactus. That is the whole list. Cactus does not train models on it, it is not sold, and it is not handed to anyone for advertising.

How long we keep it

  • Slack and Discord messages: only the ones sent to Cactus and the team's replies in a thread it handed to them, kept with the conversation they belong to and deleted when it is.

  • Conversations and the pages Cactus writes: until the owner deletes them, or closes the account.

  • What we read from a repository, a help site or Intercom: for as long as that source is connected to a product. Delete the product and it goes with it.

  • Your account name and email: until the account is closed.

  • Usage events and errors: while the account is open. Write to us and we delete them sooner.

  • Where you came from: for as long as the waitlist entry, account or team it is kept with.

Who else sees it

Nobody buys it from us. These are the companies that run parts of Cactus, and what each of them receives.

  • Cloudflare runs the website, the app and every Help Center, and keeps a log of each request.

  • Amazon Web Services runs the server where Cactus reads your sources and writes pages and answers, keeps that server's logs, and stores the database, including the search index of your code.

  • Restate runs Cactus' jobs, and keeps a record of each job's steps, including what each step read, for seven days.

  • Convex carries live updates, including the words of an answer while it is being written.

  • Modal copies your repositories, builds the search index of your code, and runs the commands Cactus reads it with.

  • The Vercel AI Gateway passes Cactus' model calls on to the company that runs each model. Calls to Anthropic and Microsoft Azure, and some calls to OpenAI, skip it and go to them directly.

  • OpenAI’s models can read your code and other sources, your pages, and your customers’ conversations, to write pages and answers. OpenAI also turns text into search vectors and draws pictures.

  • Anthropic’s Claude models can read your code and other sources, your pages, and your customers’ conversations, to write pages and answers. Every call to Claude goes to Anthropic directly.

  • Meta’s models can read your code and other sources, your pages, and your customers’ conversations, to write pages and answers.

  • Relace runs open models, such as DeepSeek’s, that can read your code and other sources, your pages, and your customers’ conversations, to write pages and answers. DeepSeek itself receives none of it.

  • TypeSafe AI’s model makes small decisions, such as whether your customers would notice a release, and reads the part of your sources or conversations each decision is about.

  • Voyage AI ranks the passages a search finds, and reads each question with the passages found for it.

  • While the Vercel AI Gateway is unavailable, Microsoft Azure turns the questions people ask, your code, and your other sources into search vectors.

  • Firecrawl reads your product's public website, to find its logo, its colours and its links, and fetches a public page for Cactus when a site turns a plain request away.

  • Resend sends our email, including the email your customers get when they write to your team.

  • PostHog records usage events, errors and recordings of visits to the app and to Help Centers, and keeps traces of model calls for debugging, which only Cactus' two founders can open.

  • Google and logo.dev supply the small logo shown beside a product's name. The browser asks logo.dev for it by the product's web address, and our server asks Google for the site's icon by the same address.

  • Google and GitHub log you in, when you choose to log in with them.

Payments are taken by Polar Software, Inc., which sells Cactus as our merchant of record. We send it your team's id and the email of the Owner who checks out. Polar collects your card, billing address and tax ID in its checkout and keeps them under its own privacy policy. We can see the name, address and tax ID you give it there. Your card is processed by Stripe for Polar, and its number never reaches us.

Your choices

  • Disconnect any source in Settings, and Cactus stops reading it.

  • Delete a product, and the pages and conversations under it go with it.

  • Close the account, and your account data goes.

  • Write to us for a copy of everything we hold about you, or to have all of it deleted.

Contact

hello@heycactus.ai. We reply within two business days, and you do not need an account to write to us.

Cactus Labs Inc.
Based in Vancouver, Canada.

How an AI client connects to Cactus, and what it can see, is on the MCP page.

What Cactus keeps in your browser is on the cookie notice. How Cactus handles personal data for a team that uses it is in the Data Processing Addendum.