Privacy policy

1. The short version

• You can see everything you shared that we still keep, exactly as you shared it, download it, and withdraw it any time. A withdrawn shared project’s content is deleted within 24 hours.
• Your sessions are private. When you open a session or make a call, what you send goes to the site’s models and to Anthropic, which runs them, so the Agent can answer. The people whose shared projects make up that Agent don’t see it.
• The site’s models read what others shared as context while answering a call; Anthropic, which runs them, receives it for that. A result can quote a lesson word for word, and no single call can hand over a large part of one contribution. Apart from files marked as starting files (templates), we don’t hand what you shared to anyone as a file. We don’t license it or sell it as data, and we don’t train or fine-tune any AI model with it.
• We collect what we need to run the site: your account, your sessions and results, your shared projects, your credits and the payment records behind them. We don’t share any of it with advertisers.

GoGoPortal is responsible for your data. We store your account, sessions, files and shared projects on our own server.

2. What we collect

• Account: your email, a name only if you choose to give one, a public username only if you choose one, and either a hashed password or, if you sign in with Google, the id Google gives us for you and the sign-in tokens Google sends with it. We ask Google only for your email address: we do not take your name or your picture from it.
• Sessions: what you are working on when you open a session, the details you send in a call, your answers to an Agent’s questions, the files you add, every result and the site’s checks on it, and any report that a result was wrong, out of date or not relevant.
• Shared projects: what you sign off to share, word for word, with its earlier versions, what the final check declined or refused and why, what each contribution earned, and the cost to learn your tool reported with it, if you kept it. While we keep it, you can see and download it.
• Credits and money: your top-ups, your credits balance, split into purchased, earned and promo, and its history, refunds to your card, and Stripe reference ids. If you switch on automatic top-up, also the amounts you set and a reference to the card you saved. Card details go to Stripe and never reach our servers. We keep a hash of the card’s fingerprint to tell when two accounts pay with the same card, because credits earned from your own calls or a linked account’s do not count.
• Connections: each connection’s name, switches and limits, its key and one-time connect code stored only as hashes, the browser or tool details of the program that claimed it, when it was claimed and last used, its requests for your approval and your answers, and a log of every call it made, refused ones included.
• Signing in from your tool: the name the program gave and the addresses it may send you back to (for a program known by its web address, also the description it publishes there); the one-time code and the tokens your tool gets, stored only as hashes; and when each was made, used, replaced or revoked.
• Technical: your IP address and browser details in server logs and for rate limiting and abuse checks, and a session cookie that keeps you signed in. Server logs can also hold your email address.
• What you send us: feedback, with the page you sent it from, your browser details and any screenshot (it is also emailed to our support inbox); reports about an Agent, with your IP address; and takedown notices and counter-notices, including any email you give.
• Decisions by staff: when someone at the site suspends or restores an account, pauses or removes an Agent, removes a contribution, decides a takedown notice, links two accounts or adjusts credits, we record who, when, and the note they wrote. The note can mention the people involved.

The site is for people 18 or older. If we learn we hold data from someone younger, we delete it.

3. Who can read a session

A session is private. What you send and what comes back can be read by:

• You, and your tool through a connection, inside the limits you set.
• The site’s models, which apply the shared research to your details to write the result. To cover what the shared research does not, a call can search the web and read pages through the model provider’s own tools; it writes those searches from your details.
• Anthropic, the model provider, which receives what you are working on when you open a session (so the Agent asks only what it still needs), your details when you make a call, and the answers you typed when a short check reads them. The privacy line under each call’s preview on the site says so, and your tool is shown it once, the first time a connection is used.
• Staff at the site, only on a report, an abuse flag, or when you ask. Each time someone opens a session, we log who, when and why.

No one else can. The people whose shared projects make up the Agent don’t see who used it, or anything from a session. A result that reaches your tool arrives marked as untrusted text, and the site shows it as plain text, never as markup.

4. Shared projects

You share what you learned, not your project. Through a connection, your tool cleans and prepares it on your computer (with the gogoportal skill or the local connector, or in Claude’s own sandbox on claude.ai), and the site refuses a batch of lessons the skill did not stamp as cleaned. That catches a tool that skips the cleaning. It can’t know your client or company names, so read what will be shared before you sign off. You can also upload a package you prepared on the site; both go through the same checks.

To check a draft before you sign off, we send it to Anthropic (a test call, and a check of any source it names). A new shared project and every update wait for your sign-off. Batches of lessons from your tool wait too, unless you choose in your Sharing settings to let them go live once they pass every check.

The site’s models apply shared material to other people’s calls. Those people pay for answers the site’s models write using it as context, and the people who shared earn credits from those answers. A result may repeat a lesson from what you shared word for word; no single call can hand over a large part of one contribution. Files you mark as starting files are copied to anyone who makes that Agent’s Starting files call, to keep, and copies already given stay after you withdraw. Nothing else you shared can be downloaded by others.

If you choose a public username and switch your public profile on, your profile shows that name, the month and year you joined, the Agents you contributed to, how many of your contributions are kept and how often your lessons were used. Your name also appears on the contributors list and on those Agents’ pages. Nothing else is shown.

You can withdraw a shared project any time; its content is deleted within 24 hours. A rights holder can send a takedown notice; the terms say what happens next.

5. Findings, corrections and the pages they name

Findings. While it answers your call, an Agent can record something it researched: a takeaway with its source, its date and the versions it applies to. It does that only if you let it. We ask once before your first call, and once for each connection. The answer is remembered, and you can change it any time on your signed-in pages. Without that choice, nothing is kept.

What goes back is the Agent’s own research, not your project. Before a finding joins, we search it for your words, your files and any term that identifies your project, and refuse it if it holds any. With it we keep one-way hashes of some of the words you sent, never the words. One that passes our checks waits for you to read it and sign off, as a shared project does. Only then does it join the Agent it fits: the one you called, another one, or a new one it starts. Findings stay when you delete the session they came from. You can see every finding from your sessions and what became of it, and remove any.

Sources.A correction and a finding each name a page: official documentation, a changelog or a release note. Our servers fetch that address, over https only, and the site’s model reads the page next to the takeaway to check it says the same thing. The request carries the address and nothing from your session. The page and the takeaway go to Anthropic for that.

6. Who else receives data

• Anthropic for the model that runs calls and the site’s own checks.
• Stripe for card payments and refunds to cards. It receives your email address and our id for your account. On the credits page, Stripe’s own script runs for payment, and it (and Google, if you pay with Google Pay) receives your device and browser details for fraud checks.
• Our email service, for receipts, sign-in links, security notices and the optional emails you choose on your account page. These can include a shared project’s title.
• Anyone the law requires us to tell.

7. How long we keep it

The keeping table awaits legal review. Where a period is not set yet it says so.

8. Your choices

You can see everything you shared that we still keep, download it, withdraw it, remove a finding, change whether an Agent may keep what it researches in your sessions, choose or remove a public username, switch your public profile on or off, choose which optional emails we send you, and delete a session. You can close your account yourself on your Account page; the keeping table says what is deleted and when. To see, correct or delete other data, write to support@gogoportal.io. We’ll act on it where the law and our record-keeping duties allow. We set only the cookies that keep you signed in. On the credits page, Stripe’s script may set its own cookies for fraud prevention.

9. Changes and contact

We may update this policy and will post the new version here, with its date. Questions go to support@gogoportal.io. See also the terms of service.