Agnoclast — legalPrivacy Policy
Effective 2026-11-12. Plain-language, and the plain meaning is the intended one. Change from the 2026-11-06 version: Section 11 now says when the full text of a reply is read from Gmail for an AI assistant you connected, and that it is not stored. Nothing that was already promised is taken away.
Changed 2026-11-06: Section 6 now says what happens when you connect an AI assistant such as Claude or ChatGPT, and Section 12 adds a support address. Nothing that was already promised is taken away.
Changed 2026-10-29: Sections 7 and 9 now describe deleting your account yourself, from Settings in the app or on the web, and exactly what that deletes and what it keeps. Nothing that was already promised is taken away.
Changed 2026-10-02: Section 5 describes a new rule for who can see captured activity. A record is visible to whoever can read the page it is filed under. Before, a scoped record stayed with its owner's management chain and teammates explicitly added to the same project, even when it was filed under a page the whole team could read. This widens what your teammates can see, so it is a material change. It applies from the outset to organizations created on or after 2026-10-02, and from 2026-10-16 to organizations created earlier — 14 days after their owner was notified — or sooner where the owner has agreed. Until then the earlier rule continues to apply in those organizations. Confidential records, records their owner has locked, and AI-session summaries captured before 2026-10-01 are not widened.
Changed 2026-09-23: Section 5 previously said scoped activity widens to teammates who "collaborate with you in the same work context." Measured against the system, that widening followed explicit project membership only — it was never inferred from sharing a repository, channel or folder, and no one is added to a project automatically. Section 5 was corrected to say so. That narrowed what we claimed was shared; nobody gained access they did not have.
Changed 2026-09-20: Section 5 now says tenant isolation is tested automatically on every merge to main, rather than on every code change. The test that runs against the production database runs once per merge; a pull request is checked against a separate copy of the database instead. Nothing about who may reach your data changed — this corrects how often we said we verify it.
Changed 2026-09-18: Sections 5 and 8 began describing operator access as it works today. People who operate the Service can reach Customer Content directly in the database, for the purposes listed there, and only part of that access is recorded and shown to your organization.
Changed 2026-09-17: Section 11 began describing what Agnoclast can do inside a connected Google account when you use its email and calendar features — mark, archive and label email, create reply drafts, send an email you approved, and add calendar events — and how a requested email is stored while it waits for your approval.
Changed 2026-09-10: Section 3 began stating when part of a session transcript reaches our servers, and Section 11 began stating how data received from Google (Gmail and Google Calendar) is accessed, used, stored, shared and deleted. Contact moved to Section 12.
Changed 2026-09-08: the controller is now named. It is Agnoclast LLC, a Utah limited liability company, which replaced Theron Peterson as operator of the Service on that date. No data practice described below changed.
1. Scope
Agnoclast is operated by Agnoclast LLC, a Utah limited liability company ("we," "us"). It is a business tool used by organizations. For content your organization puts into Agnoclast ("Customer Content" — work records, pages, summaries), your organization is the controller and we process it on the organization's behalf under our agreement with them. For the basic account data we need to run the Service (your name, email, login), we are the controller. Questions about what your employer captures should also go to your organization's admin — see the notice they provided at onboarding.
2. What we collect
- Account data: name, email, password hash (via our auth provider), role, org membership, personal access token.
- Work-activity records: when your AI session ends, a summary of what was worked on — accomplishments, open threads, decisions, people/projects mentioned. Not the raw transcript (see Section 3).
- Connected-source records: if you or your org connect sources (GitHub, Slack, email, calendar, files), work records derived from them — e.g. commit summaries, channel digests, file metadata and content fingerprints.
- Authored pages: the wiki pages you and your AI write into your org's knowledge base.
- Held (unassigned) source records: records from a connected source that has no routing rule yet are held outside every organization until that source is assigned to a workspace. While held they are readable only by the individual whose connected account produced them — not by their organization, their management chain, or their teammates.
- Operational data: sync/error logs, usage metering (e.g. AI token counts per org), audit logs of staff/operator access, and the technical basics any web service receives (IP, browser type, timestamps). We do not run third-party advertising or cross-site tracking, and we use cookies/local storage only to keep you signed in.
3. What we deliberately do NOT collect
- Raw AI-session transcripts normally stay on your machine. Summarization runs locally on the Member's own device and AI subscription, and the derived summary is what reaches our servers. When local summarization is unavailable, the most recent part of the session — with detected secrets redacted — is sent to our servers so that our AI inference subprocessor can produce the summary.
- We do not record your screen, keystrokes, or browsing.
- We do not copy file contents off your machine — file capture stores summaries and fingerprints.
- We do not capture personal accounts you haven't connected.
4. How we use information
Solely to provide and operate the Service: building your org's knowledge base, serving permission-scoped context, security, debugging, and (if you contact us) support. We do not sell personal information, do not use Customer Content for advertising, and do not use it to train AI models — and our AI subprocessor is contractually barred from training on it too.
5. Who can see what
Inside your organization, visibility is decided by pages. Every page has an owner, and the owner chooses who can read it: everyone in the organization (accessible); a part of the org chart relative to the owner — everyone below them, their direct reports, or their peers — together with the owner's managers; the owner and their managers only (scoped); or the owner alone (confidential).
Captured activity — commits, meetings, AI-session summaries and other records — is filed under pages, and a record is visible to whoever can read the page it is filed under. In an organization created before 2026-10-02 this takes effect on 2026-10-16, or earlier if the organization's owner agreed; until then a scoped record there is visible only to its owner's management chain and to teammates explicitly added to the same project. Three things stay narrower than their page:
- Confidential records never widen. They stay owner-and-grants-only, which is the tier to use for anything you want kept to yourself.
- A record its owner has locked — by setting it to scoped themselves, or by asking for scoped when it was created — stays with its owner and their management chain.
- AI-session summaries captured before 2026-10-01 are treated as locked until their owner releases them. Summaries captured from that date follow their page like any other record.
Separately, and as before, a scoped record that belongs to a project is visible to teammates who have been explicitly added to that project, whatever page it is filed under. Project membership is never inferred from sharing a repository, channel, or folder, and no one is added to a project automatically. Email and iMessage records are visible only to the people on the message, and calendar events also to the people invited. Activity not yet filed under any page is visible only to the person it belongs to, or to the organization's owner and admins when it belongs to no one. You can lock or release any record of yours at any time. Across organizations, tenant isolation is enforced at the database layer and tested automatically on every merge to main. People who operate the Service can also reach Customer Content directly in the database, outside these controls. They do so only to run, maintain, secure and support the Service, when you ask them to, or when the law requires it, and for data received from Google the stricter limits in Section 11 apply. Some of that access is recorded and shown to your organization in the console's Data-access panel, and some is not yet. Full detail, including which is which: Data Access & Tenant Isolation.
6. Sharing
We share data only with the infrastructure subprocessors needed to run the Service (hosting, database, AI inference) — the current list, and what each one sees, is at /legal/subprocessors — and when required by law (we will notify your organization of legal demands unless legally barred). No sale, no ad networks, no data brokers.
AI assistants you connect. When you connect an AI assistant such as Claude or ChatGPT, it receives what you ask it to read from Agnoclast: your pages, obligations, calendar and records you can see, including email headers and message text from sources you connected. That provider handles it under its own terms. Some may use it to improve their models unless you turn that off (in ChatGPT: Settings → Data controls → Improve the model for everyone). You can disconnect an assistant at any time from Connect in Agnoclast or from the assistant's own settings.
7. Retention and deletion
- Customer Content is retained while your organization's account is active.
- Held (unassigned) source records are retained until the source is assigned to a workspace — at which point they become Customer Content of that organization — or until the individual deletes them or disconnects the source. They are not silently discarded, and they are not attributed to any organization while held.
- Members can re-tier or request correction/deletion of their records via their org admin or the in-product controls.
- On organization termination: export available for 30 days, then deletion from production; encrypted backups age out on the backup provider's schedule (typically ≤30 additional days).
- Deleting your account (Settings → Delete account, in the app or on the web) takes effect at once. It deletes your sign-in and the personal data tied to it (your settings, tokens, devices, sessions and billing details), deletes the records you captured, disconnects every source you connected, sets every subscription you pay for to end at the close of its current period, and removes you from every workspace. Where you owned a workspace that other people use, ownership passes to the successor you named or to an admin. What is kept: earlier versions of pages and records are kept in a locked, append-only history that only our operators can reach. Like the workspaces themselves, it cannot be deleted. Your name and email are removed from your membership records, so that history is no longer linked to you. Encrypted backups age out on the backup provider's schedule (typically ≤30 days).
- Operational logs are kept for security and debugging and pruned on a rolling basis.
8. Security
Postgres row-level security enforces tenant and tier isolation on every read; an automated security test suite runs against the production database on every code change; credentials are rotated; and operator access is purpose-limited and, where Section 5 says so, recorded and shown to your organization. No internet service can promise zero risk, and we don't — see the honest limits in Data Access & Tenant Isolation. If we learn of a breach affecting your data, we will notify your organization without undue delay.
9. Your rights
Depending on where you live you may have rights to access, correct, delete, or export your personal data. For Customer Content, direct requests to your organization's admin (they control the tenant); for account data, delete your account yourself from Settings (Section 7), or contact us directly and we'll respond within 30 days. We honor these rights regardless of jurisdiction where practicable.
10. Miscellany
- Location: the Service is operated from the United States and data is stored there. By using it you understand your data is processed in the US.
- Children: the Service is for working adults; it is not directed at anyone under 16.
- Changes: material changes to this policy are notified to organization owners before taking effect; the current version always lives at /legal/privacy.
11. Google user data (Gmail and Google Calendar)
This section applies when you connect a Google account. Where it differs from the rest of this policy, this section controls for data received from Google APIs.
- What we access. Gmail: the mailbox you connect — each message's sender, recipients, date, subject and content, and its labels (for example read or unread, and Inbox). Google Calendar: your calendar events — title, time, attendees, location and description — and the email address of the Google account you connected.
- What we do in your Google account. Only what you ask for, through the features you use: - Email: mark messages read or unread, archive them, add or remove labels, create reply drafts for you to review in Gmail, and send an email. An email is sent only after you approve that exact message in Agnoclast in your browser, or under a sending rule you set yourself there. An AI assistant connected to Agnoclast can ask to send an email, but it cannot approve its own request. - Calendar: add an event for a deadline when you choose "Add to Google Calendar". Agnoclast may also add an event for a deadline it has identified as yours with high confidence; when it is unsure, it asks you first. - Never: we do not delete your email, and we do not edit or delete calendar events you created. Agnoclast only updates or checks events it added itself.
- How we use it. Only to provide the features you connected it for: turning messages and events into records in the brain you choose, serving those records back to you — and to the people you have given access under Section 5 — through Agnoclast and the AI assistant you connect to it, and carrying out the actions described above.
- Emails waiting for approval. When an email is requested, its recipients, subject, body and attachment references are stored with the request, so that the message you approve is exactly the message sent. That stored copy is deleted 7 days after the request is created, whether or not it was sent. An email that is sent also appears in your Gmail Sent folder and is captured like your other mail.
- Replies to email your assistant sent. When an AI assistant connected to your Agnoclast account sent an email for you and someone replies on that thread, the assistant can open the reply and see its full text, so it can answer what the person actually said. That text is read from Gmail at the moment the assistant opens that one reply, only for you as the owner of the mailbox, and only on a thread your own assistant sent on. It is shown to the assistant for that view and is not stored by Agnoclast. Earlier messages quoted in the reply are left out, and very long replies are cut short. Otherwise, ongoing capture of your mail keeps only each message's details and Gmail's short preview.
- Where it is stored. Google access tokens are encrypted at rest — in our database when you connect on the web, or in your device's secure credential store when you connect through the Agnoclast desktop agent. Records created from Google data are stored with the rest of your Customer Content (Section 7), in the United States.
- Who it is shared with. Only the infrastructure subprocessors listed at /legal/subprocessors, only as needed to run these features, and when required by law (Section 6). Google user data is never sold, never transferred to advertising platforms, data brokers or information resellers, never used to serve advertising, and never used to determine creditworthiness or for lending purposes.
- No AI training. We do not use Google user data to develop, improve or train generalized AI or machine-learning models, and our subprocessors may not either.
- No human reading. Agnoclast personnel do not read your Google user data unless you ask us to for a specific item (for example, for support), it is necessary for security purposes such as investigating abuse, it is required to comply with applicable law, or it has been aggregated and anonymized for internal operations. Operator access is otherwise limited as described in Section 5.
- Revoking access and deleting data. Disconnect the source on the Data sources page in Agnoclast, or remove Agnoclast at myaccount.google.com/permissions. Either stops new capture. To delete records already created, use the in-product controls, ask your organization's admin (Section 7), or contact us (Section 12).
- Limited Use. Agnoclast's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
12. Contact
Privacy questions or requests: theronpeterson654@gmail.com.
Help using Agnoclast: support@agnoclast.com, or agnoclast.com/support.