ThreadPool

Last updated 24 September 2026

Privacy Policy

This policy is for ThreadPool, a product of YourAmaryllis (youramaryllis.us). It covers the macOS, Windows, iPhone, and Android apps and any later clients that talk to the same service. It is the listing policy for Slack, GitHub, Atlassian, and Google. It is not a substitute for counsel.

What ThreadPool is

ThreadPool clusters your work (messages, issues, pull requests, mail, and docs) into topics and can show prior decisions during meetings. It is not a recording archive and does not join calls as a bot.

Who we are

YourAmaryllis operates ThreadPool. Questions: support@youramaryllis.us or the contact page.

What we collect

ThreadPool is a work aggregator. To cluster topics we ingest work from services you connect. Topic summaries, statements, and OAuth tokens are encrypted on your devices before they are stored. We keep ciphertext and metadata for longer than 24 hours. That includes:

  • Account: email via sign-in. We map that to an internal user id. Device public keys so sibling apps can unwrap a space key. A recovery backup is wrapped with a secret you save once. We cannot unwrap it.
  • Connected services (you click Connect): OAuth tokens for Slack, GitHub, GitLab, Jira, Linear, Miro, Google, and API credentials for Lightfield. Tokens are encrypted with your space key on the client. Ciphertext is stored on our servers, scoped to your account and space. Company Client Secrets never ship in an app, Keychain Settings fields, or logs. Vendor identity after Connect: Slack user id, GitHub login, Atlassian accountId, display name, and email when the vendor gives them to us.
  • IMAP: host, username, and port metadata may be stored so the desktop can reconnect. The mailbox password stays in the OS keychain on that device. We do not upload it.
  • Work graph: encrypted summaries of threads we ingest, plus metadata we need to cluster and retrieve: people maps, source links, which Slack workspace, GitHub or GitLab projects, Jira site and projects, Linear teams, Miro boards, Google sources, and Lightfield objects you chose. We do not need your GitHub or GitLab source code (PR description, comments, commit messages only) or Jira attachments.
  • Meetings (opt-in): calendar title and attendees for pre-brief. Live capture is captions-first or on-device audio. We do not send meeting audio to our cloud. We do not add a participant bot to Zoom, Meet, or Teams. Companion cards on a phone are session metadata and cards, not the raw transcript.
  • Bring-your-own AI: provider choice may be stored. The API key stays in the OS keychain. We do not upload it.
  • Billing: if you subscribe, Stripe sees your payment method. We store Stripe customer and subscription ids, not card numbers.

We do not sell personal data. We do not use your work graph to train public models. We cannot read encrypted topics or tokens without a key that lives on your devices (or your recovery secret).

What we do not pull

  • Every Slack channel you can read, only places you already spoke in.
  • GitHub or GitLab source code, diffs, or file contents.
  • Jira attachment bytes, or every issue in a project you did not pick.
  • Linear issues outside the teams you picked.
  • Miro comments. Their API does not provide them.
  • Gmail attachments or Drive file bytes.
  • IMAP passwords, BYO AI keys, or meeting audio.

Google user data

When you connect Google, ThreadPool accesses only the Google user data needed to cluster your work. We read:

  • Account: the email, name, and profile of the Google account you connect, so we know who granted access.
  • Gmail: message text from your Primary inbox and starred mail from the last 14 days. We do not download attachments. We do not send, delete, or label mail.
  • Google Chat: messages from the last 14 days in spaces where you sent a message or were mentioned. We do not post messages.
  • Google Drive, Docs, Sheets, and Slides: files of those types that changed in the last 14 days, comments on them, and the paragraph, row, or slide text around a comment so the comment can be clustered. We do not download file bytes. We do not create, edit, or delete files.

We use that Google user data only to provide ThreadPool: to summarize those threads and comments, cluster them into your topics, and show prior decisions when you ask. We store the resulting summaries and the OAuth tokens encrypted on your devices before they reach our servers. We keep that ciphertext until you disconnect Google or delete your account, then we delete the Google tokens and the Google-derived rows. You can also email support@youramaryllis.us and ask us to delete them.

We share Google user data only with processors that run that feature: Supabase stores the encrypted rows, and a model provider receives a topic card only when you ask ThreadPool to summarize or answer. We do not sell Google user data. We do not transfer or disclose it for advertising, to data brokers or information resellers, or to determine credit-worthiness or for lending. We do not transfer it to anyone for a purpose other than providing or improving ThreadPool’s user-facing features.

The use of information received from Google APIs, including Google Workspace APIs, will adhere to the Google API Services User Data Policy, including the Limited Use requirements. We do not use data obtained from Google Workspace APIs — raw or derived — to develop, improve, or train non-personalized AI or machine-learning models. If we change how ThreadPool accesses, uses, stores, or shares Google user data, we will update this page and say so in the app.

Who else sees it

We use processors to run the product: Supabase (auth and encrypted database rows), Google Cloud (company-provided models when you have not brought a key), Stripe (payments), and the vendors you connect (Slack, GitHub, GitLab, Atlassian, Linear, Miro, Google, Lightfield, IMAP hosts you name). Each sees only what that job needs.

If you use company-provided models for Ask or HUD judging, retrieved topic cards are sent to that model to produce an answer. If you bring your own AI key (Gemini, Claude, Groq, Grok, NVIDIA, OpenAI), that generate step runs with your contract. Embeddings for clustering use a consistent company model so topics stay comparable.

Because we keep Atlassian account identifiers, display names, and emails, we report those accountIds to Atlassian about once a week through their Personal Data Reporting API. If Atlassian says an account is closed or the profile changed, we erase or refresh our copy. We do not send your Slack or GitHub data in that report.

We disclose if required by law.

How long we keep it

Until you disconnect that integration, delete your account, or ask us to erase it. Disconnecting removes that vendor’s tokens and stops new pulls. A full account delete (Settings or email us) removes your ThreadPool rows (topics, people, credentials, devices, billing customer mapping). Stripe may retain invoices as required by tax law. Backups roll off on our host’s schedule. Email support@youramaryllis.us if the in-app controls are not enough.

Your choices

  • Do not connect a vendor you do not want us to read.
  • Pick specific GitHub repos, GitLab projects, Jira projects, Linear teams, and Miro boards.
  • Use a separate space for Work vs Family so those graphs never mix.
  • Bring your own AI key if you do not want company models to judge Ask.
  • Disconnect or request deletion at any time.
  • If you are in the EEA/UK or California: access, correction, deletion, export. We do not sell or share for cross-context ads. You may also complain to a supervisory authority.

Children

Not directed at children under 16.

Changes

We will post a new date on this page. Material changes to how we use Google user data, to meeting capture, to encryption, or to subprocessors will be called out in the app.