All docs

Zero conversation logs

We do not store what you send or what the model answers. This page lists exactly what is kept, for how long, and what the promise does not cover.

What is not stored

  • Prompts and context are not saved
  • Documents and codebases are not scanned
  • Request content is not written to logs

Prompts and outputs pass through our systems only so they can be forwarded to the model and the response returned to you. We do not use them to train models.

The API keeps no conversation state on the server either. On the Responses endpoint, store and previous_response_id are not supported: your client sends the full input each turn.

What is stored

Running a paid API needs some records. These are the ones we keep:

Account
Email address, username, and your password as an Argon2id hash.
Usage metadata
For each request: account and key ID, model, token counts, cost, timestamps, client IP address, request ID, status and error information. No prompt or completion content.
Billing
Top-up records: order number, amount, payment method, status and time. Card payments are handled by Stripe; we never receive full card numbers.
Security
IP addresses, login sessions and CAPTCHA results.
Visit statistics
Hashed anonymous visitor and visit IDs, country, browser language, referrer and landing path. No IP address or user agent.

How long it is kept

Usage and billing records
1 year, as needed for accounting and tax.
Application log files
Deleted after 14 days.
Account data
While the account exists, and up to 30 days after deletion.
Database backups
Encrypted, overwritten after roughly two to four weeks.

The application server and database are in Germany (EU).

What the promise does not cover

We would rather say this plainly than have you find it in the fine print:

  • The model provider sees your prompt. To generate a response, the request goes to whoever serves the model: a third-party inference service or a server we run. A third-party service handles it under its own terms, not ours. Email support@6bee.ai to learn who serves a particular model.
  • Metadata is logged. We know that an account made a request, when, to which model and how many tokens it used. We do not know what it said.
  • Your side is yours. Your client, your shell history and your editor may keep their own copies.

What you send is up to you. Keep your own prompts and code safe.

You can check the code

The gateway and billing code that runs the console and API is public under AGPL-3.0. The exact version in production, with our patch and build steps, is on the source page.

Questions

Can 6bee.ai staff read my conversations?

There is nothing stored to read. Request content is not saved and not written to logs.

Do you scan code or documents I send?

No. Documents and codebases are not scanned.

Can I use the API without giving my identity?

Yes. Sign-up needs only an email address, and you can pay in USDT or USDC. We collect no phone number, no identity documents and run no KYC checks. Per-request metadata still includes the client IP address.