Privacy Policy

Dylan Davies, trading as Jugg.ai, Wellington, New Zealand, is the agency (under the New Zealand Privacy Act 2020) and the data controller (under the UK/EU GDPR) for the personal information described here.

Privacy Officer: Dylan Davies — privacy@jugg.ai.

• Account. Your name, email address and avatar as your sign-in provider (Google, GitHub or Microsoft) reports them, and the OAuth tokens for that sign-in — encrypted at rest with AES-256-GCM.
• Organisation. Org name, members and their roles, invitations, plan and subscription state, and your spend caps.
• Model provider keys. The API key you paste for your model provider, encrypted at rest. It is never returned to the browser and never logged — the product only ever shows a masked hint.
• Connections. For each tool you connect: its URL, its scopes, and its credentials or OAuth tokens, encrypted at rest.
• Telegram. If your organisation connects its own Telegram bot: the bot’s name and token (encrypted at rest), and for each person who links a chat, the Telegram chat id that links it to them. The messages you send the bot become jobs like any other; Telegram itself handles them under its own privacy policy.
• Slack. If your organisation connects its own Slack app: its bot token and signing secret (encrypted at rest), the workspace and bot ids, for each person who links, their Slack user id, and for each job started from Slack, the conversation and thread it belongs to. The bot only receives what is said to it — mentions and direct messages — and those become jobs, answers or guidance. Slack handles them under its own privacy policy.
• Work that arrives. If a team has its own email address or webhook URL, what is sent to it — an email’s sender, subject and text, or a webhook’s body — becomes the input to a job and is kept with that job. Attachments are never fetched.
• Web pages a team reads. When a team opens a public web page for a job, our servers request it from that site (identifying as JuggBot and following the site’s robots.txt). The site sees our server, not you. The page’s text goes into the job like any other result; we keep only a one-way fingerprint of each address the job may open next, deleted with the job.
• Work. The jobs you submit, the missions and skills you write, the task board, agent activity logs, the reports your agents produce, and estimated token usage per call.
• The Hive. Your teams’ shared memory — the findings, reports and facts your agents write, plus embeddings computed from them for search.
• Docs. The documents you (or your agents, as drafts) write in Docs, their revision history, and the search index built from them (text sections, plus embeddings if you add an OpenAI key). Files in your connected storage — Google Drive, OneDrive, SharePoint — stay there: we do not copy or store them. An agent opens one only when its task needs it, through the connection you granted, and what it reads is handled like any other tool result (see Work, above).
• Audit log. Who did what in your organisation, and what your agents did on your behalf.

Google user data. If you connect Gmail, Google Calendar or Google Drive, your agents use that access only to do the work you give them: sending the emails your jobs call for, reading and editing calendar events, and opening or creating the Drive files your jobs use. Content read through these connections goes to the AI model provider you chose, under your own key, only as part of the job that needs it. It is never used to train or improve any AI model, never sold, and never used for advertising. People never read it unless you ask us to (for support), we need to for security or abuse investigations, or the law requires it. Jugg.ai’s use and transfer of information received from Google APIs to any other app will adhere to the Google API Services User Data Policy, including the Limited Use requirements. You can disconnect at any time from Connections on the dashboard, or revoke access from your Google Account.

Whatever you or your agents put into a job, a mission or the Hive is stored as written. If that includes someone else’s personal information, you are responsible for having a lawful basis to put it there.

• To perform our contract with you — running your teams, remembering what they learn, enforcing your caps, billing your subscription.
• Legitimate interests — keeping the Service secure and available, preventing abuse, debugging failures, and understanding aggregate usage.
• Legal obligation — tax and accounting records for payments.

We do not train models on your data, we do not sell it, and we do not share it with advertisers. There is no ad tracking anywhere in the product.

We use these providers to run the Service. Each is bound by its own agreement with us and may only process your data on our instructions.

The model provider is the one you choose, and it is the one that sees the substance of your work. Settings → Model provider states, for each provider, which country processes your prompts — read it before you connect customer data. Because these providers are overseas, using the Service involves a cross-border disclosure of personal information under Information Privacy Principle 12.

The 3D views in the product load font data for their labels from a public CDN (cdn.jsdelivr.net). That request comes from your browser, carries no account or work data, and reveals only your IP address and browser details to the CDN, as any web request does.

We will update this list before adding a subprocessor that handles customer data.

• Account, organisation and work — for as long as your organisation exists. An owner or admin can export it, and an owner can delete the organisation, at any time from Settings → Your data (see §11 of the Terms); deletion removes every row keyed to it by database cascade, immediately. Your billing record at Stripe (invoices and the customer they belong to) is kept, because tax law requires it — see Billing records below.
• Audit log (who did what: approvals, decisions, changes to settings, teams and connections) — for as long as your organisation exists, and exported with it.
• Agent activity logs (each step an agent took) — 30 days, then removed automatically by a nightly retention sweep.
• Support tickets — while open, and for 12 months after a ticket is closed (and nobody has written on it since); then the ticket and every message in it are removed automatically by the nightly retention sweep. A record that our support staff opened, replied to or changed a ticket (who, which ticket, when — never what it said) is kept as a security record for two years. A ticket waiting on your reply for 30 days is closed automatically (replying re-opens it).
• Email to our support, privacy or security addresses — we keep the plain-text part only (never HTML or attachments). Our support staff may turn it into a ticket for the organisation it concerns; otherwise it is deleted 90 days after they dismiss it, or after 180 days if it is never handled.
• The Hive — kept deliberately: it is the memory that makes a standing team worth having. The Hive is retained for as long as your organisation is active. An owner or admin can delete any entry from the team’s Memory tab at any time, or ask us at support@jugg.ai to delete specific entries or the whole Hive. A nightly sweep re-embeds entries when the search model changes, so their text is sent to the embedding model provider again at that point (the embeddings run on your OpenAI key, under “Your chosen model provider” above).
• Billing records — kept for as long as tax law requires, typically seven years.
• Credentials — deleted when you remove the connection or the provider key.
• Telegram and Slack links — deleted when you unlink, or when the bot or app is disconnected (a job’s Slack thread reference goes with it).

You may ask us to give you a copy of the personal information we hold about you, correct it, or delete it. Under the GDPR you may also object to or restrict processing, and ask for portability. An organisation’s owner can export its data (a portable JSON file) or delete it at any time from Settings → Your data. For anything else, email support@jugg.ai — we aim to answer within a few business days and, for a formal request under the Privacy Act, within the statutory 20 working days.

If you are not satisfied with how we handle it, you may complain to the New Zealand Office of the Privacy Commissioner or, in the EU/UK, to your local supervisory authority.

Every tenant-keyed table is protected by database row-level security, so one organisation’s queries cannot reach another’s rows. Provider keys and connection credentials are encrypted at rest with AES-256-GCM. Custom tools your agents write run in a sandbox with no ambient network or filesystem access. Secrets are scrubbed from logs.

If a breach occurs that is likely to cause you serious harm, we will notify you and the Office of the Privacy Commissioner as required by the Privacy Act 2020, without undue delay.

Found a vulnerability? Please report it to security@jugg.ai rather than disclosing it publicly. We will not pursue researchers acting in good faith. We hold no security certifications — we are not SOC 2 audited, and we will never imply otherwise.

We set only what the product needs to work: a session cookie so you stay signed in, and a cookie recording which organisation you are currently acting in. There are no advertising cookies, no third-party trackers and no cross-site profiling, which is why you are not being asked to dismiss a banner. To see how many people visit, we use Vercel Web Analytics, which counts page views and a few steps (a sign-up, a first team, a first job) without cookies, without identifying you across sites, and never with what you or your teams write.

The Service is for business use and is not directed at anyone under 16. We do not knowingly collect their personal information.

We will update this page when our practices change, and change the “Last updated” date above. For a material change affecting how we use your personal information, we will tell you by email.