Privacy at Starpod

Updated on October 19, 2026

For any question about your information (knowing what we have about you, getting a copy, having something corrected or deleted), write to Starpod’s privacy officer at privacy@starpod.ai. The contact@starpod.ai and privacy@starpod.ai inboxes are read by a person on our team, with an AI assistant that sorts them: your email therefore goes through Claude, by Anthropic, in the United States. A privacy request never gets an automatic reply: a person on our team answers it. It is free, and we answer in writing within 30 days.

Our full privacy policy is on its way: it will name each of our providers, with where they process your information, and how long we keep each thing. We would rather publish it once every fact is checked than show you one that is half true.

Until then, what is already true: we never sell your information, and we never use it for advertising. And if our answer does not satisfy you, you can turn to the Commission d’accès à l’information du Québec.

Fireworks AI, an American company, processes your messages for us. To produce each reply, to summarize a long conversation and to check the replies, it receives the conversation (what you write, the earlier messages and replies, and the instructions we give the assistant), the contents of the files your pod reads or writes for your work, the results of the commands it runs, and the images you send.

Three other American companies work for us. Stripe handles payments: it receives your email address and our reference for your account, and you type your card on its page, never on ours. Resend sends the emails we send you: it receives your email address and the content of each email. Linode hosts our servers, your pod included.

Our team is made of people and AI agents, and they may read some of what reaches them, for security and quality. Our AI agents run on Claude, by Anthropic, in the United States. What can reach the team: the short notes your pod sends when it needs help (the kind of work, the number of attempts and a reason of 160 characters at most, written by the pod’s agent), what you choose to send to the team by email, and the transcripts of our AI agents’ work sessions, which may contain your words. We do not enter your pod to read your data. The administrator account of the server could technically read it; your panel lists these ways in and says which ones leave no record.

The chat and its database do not keep your IP address as is, but a fingerprint: a code derived from the address with a secret key on our server. These fingerprints are deleted after 90 days. Backup copies of the database are not rewritten: they keep them until they rotate out. Our nightly backups are encrypted on our own server, with a key Proton does not hold, and stored with Proton.

One log does not leave with your account: the usage log. For each reply from the chat, it keeps the time, the model used, the size of the exchange and its cost, with the fingerprint of your IP address. It holds no message, and nothing in it names your account. Each line is deleted after 90 days.

Our web server’s access log keeps your address truncated (its last part is removed), for 14 days. Before September 24, 2026, it kept it whole; those old logs were all deleted from the server by October 10, 2026 at the latest. The error log keeps the whole address of a request that failed, for 49 hours at most, since October 5, 2026; the earlier ones, kept for 14 days, were all deleted from the server by October 7, 2026 at the latest. Our host (Linode) also keeps a full copy of the server, replaced within 16 days at most: these logs stay there until then.