What we store, and what we don't
Your brief, your material and your feedback stay. Screenshots and social logins don't.
You are about to tell a piece of software how you think and let it post under your name. Here is exactly what it keeps, what it throws away, who else touches it, and how to make it all disappear. The Privacy Policy is the formal version; this is the useful one.
What we keep
| Thing | Kept? | Why, and for how long |
|---|---|---|
| Your answers to the setup conversation | Yes | They are the agent's brief. Kept while the agent exists; editable at any time. |
| Text read from your LinkedIn screenshots or profile PDF | Yes | Stored as two notes attached to your agent — "My past LinkedIn posts" and "My LinkedIn profile" — that you can read, edit or delete like any other source. |
| The screenshots and PDF themselves | No | Read once, in memory, to extract their text. Not written to storage. This is why a resumed setup cannot show them again. |
| Documents you add as material | Yes | Stored privately in cloud storage with no public access. Deleted when you delete the source. |
| Generated posts, images, your thumbs and notes | Yes | They are your history and the agent's examples. Images live in private storage and are served through short-lived signed links. |
| Your social media password | Never | You sign in on the platform's own page. We never see it. |
| Authorization tokens for connected accounts | Held by our publishing partner | The token that lets a post be published is issued by the platform to our publishing partner and stays there. Revocable from your platform's settings at any time. |
| Your card details | No | Stripe holds them. We keep your subscription status and the last four digits. |
What we never do
- Scrape your LinkedIn. Connecting an account grants permission to post, not to read your profile, your posts or your network. What the agent knows about your history is what you gave it.
- Train shared models on your material. Your brief, your documents and your posts are used for your agents and no one else's.
- Pre-fill your voice from your profile. A profile has no posts in it, so the agent leaves the voice question empty rather than guess. Only real samples fill it.
- Publish something you removed. A post pulled from the outbox becomes a draft and stays one.
Who else sees your data, and what part
- The writing model (Anthropic) receives your brief, the material fetched for that post, and your recent graded posts — every time a post is written. Its terms do not permit training on this data.
- The image model (OpenAI) receives only the one-sentence image concept, never your documents or brief.
- Our publishing partner (Zernio) receives each finished post and image at publish time, and holds the authorization tokens for your accounts.
- Our analytics (PostHog) logs the prompts sent to the writing model and the responses, so we can watch quality and debug problems. This means your material appears in those logs. They are visible to our team only, and we will delete them on request.
- Our email provider (Resend) receives the address and content of each email we send you.
Agencies and clients
A client's data belongs to the agency's organization. It is isolated from every other client and from the agency owner's personal agents: each client has their own space with our publishing partner, holds exactly one account per platform, and a client's agent can only ever post to that client's accounts. A client without a login sees only what the agency chooses to send them, through a private link that expires after 30 days. See What your client sees.
Deleting everything
Deleting your account from Settings removes your agents, brief, material, posts, images, connected accounts and the space at our publishing partner, cancels any subscription, and — if you own an agency — does the same for every client. Backups clear within 30 days. Billing records are kept as long as the law requires. If you want the analytics logs cleared as well, email support@thepostagent.com and say so.