Privacy policy

Data categories

Account: your Clerk user id, which is the key for your workspace. Your email and login live in Clerk, not in a separate seodraft copy. The one exception is early access granted before signup, which is recorded as an allowlist entry keyed by the email address you were invited with.

Business profile: business name, description, audience, competitors, voice, site URL, sitemap URL, reference URLs, and your writing-style preferences.

Evidence bank: the first-hand facts you or your agent add for articles (a measured number, a named case, a mistake, a credential), each marked pending or accepted.

Topics and metrics: the search topics you save, plus any Google or AI search volume, difficulty, and intent numbers measured through your own DataForSEO account or observed by a site audit.

Drafts and briefs: article bodies, editorial briefs, SERP research briefs, and the results of the deterministic checks (the rules) run against each draft.

Your site's sitemap: an inventory of your own site's published pages, built from your sitemap.xml.

Site audit: what an audit of your site observed, stored for your workspace: the public pages of your site with their title, headings, description, status code, word count and whether they can be indexed, the internal links between them, the search terms your site and the competitor domains it names rank for, and which AI answers cite your site. All of it is public information about websites, none of it personal data.

DataForSEO usage log: a record of each paid call made for your workspace, whether against your own DataForSEO account or as part of a site audit seodraft funded (action, cost, timestamp, and who paid).

Delivery target: the git repository, branch, file-path template, and format you configure for exporting drafts.

Credentials, encrypted: your DataForSEO login and password, and your GitHub token. See "Your DataForSEO and GitHub credentials" below.

Activation events: internal counters for onboarding steps (agent connected, first brief written, and similar milestones): ids and timestamps, never article text.

Transactional email log: which emails we sent you (welcome, trial reminders, and similar) and delivery events our email provider reports back (bounced, opened).

What each category is used for

Account: identifies your workspace and lets you sign in. The allowlist entry exists only so an invited address can create an account.

Business profile: tells your agent who it writes for, in which language and voice, and which site to link into.

Evidence bank: feeds the rule that requires first-hand facts in every article, and is what your agent cites while drafting.

Topics and metrics: rank what to write next, and feed the check that stops two articles from chasing the same query.

Drafts and briefs: are the articles themselves, the plan behind each one, and the record of which rules passed.

Your site's sitemap: avoids proposing a topic a published page of yours already covers, and supplies internal link targets.

Site audit: shows the Map, and is the evidence your agent cites when it proposes clusters of pages and topics.

DataForSEO usage log: shows what your own DataForSEO account was charged for, keeps the site audits seodraft funded apart from that spend, and enforces the spend caps.

Delivery target: tells seodraft which repository, branch and path to commit an approved draft to.

Credentials: authenticate the calls you or your agent trigger against DataForSEO and GitHub, and nothing else.

Activation events: tell us which onboarding step a workspace stalled on, so the product and the reminder emails can name it.

Transactional email log: records which emails were sent, and stops the same one being sent twice.

Your DataForSEO and GitHub credentials

Your DataForSEO login and password, and your GitHub token, belong to you, not to seodraft. Every topic measurement and every brief your agent requests runs against your own DataForSEO account and is billed to your own balance. The one exception is the site audit behind the Map, including the briefs it builds after you approve a map: seodraft may run it through a DataForSEO account it operates, and then sends DataForSEO only public information (your site's domain, competitor domains, the search phrases being researched, and the workspace's language and country), never your credentials or personal data. With your own DataForSEO account connected, the audit runs on your account instead. Your credentials and token are stored encrypted at rest with AES-256-GCM, decrypted only at the moment a call you or your agent triggered needs them, and used for nothing else.

No MCP tool returns your DataForSEO password or your GitHub token. Tools that report on delivery or DataForSEO status return whether a credential is connected, never the credential itself.

Who else processes your data

Authentication: Clerk, which holds your login, your email address and your session.

Hosting and database: Vercel hosts the application; Neon runs the Postgres database everything above is stored in.

SEO data: DataForSEO, for search volume, difficulty, and SERP data, called with your own credentials. The one exception is a site audit seodraft funds, which runs on seodraft's own DataForSEO account.

Content delivery: GitHub, only if you configure a delivery target, and only with the token you provide, to commit draft files to a repository you choose.

Billing: Commet, for the subscription.

Email: Resend, for transactional email (welcome, trial reminders, and similar).

Analytics: Umami, privacy-focused analytics on the public marketing pages only (this page, /pricing, /blog, and similar). It is not loaded on your workspace pages, so it never sees a post id, a draft, or your profile.

Classification: TypeSafe Jev, an optional classifier that runs only when seodraft has it enabled on the server. It answers typed questions about a topic, a draft, or a profile field (for example, whether two topics are semantic twins, or whether a piece of evidence reads as first-hand) and returns a probability. It never blocks anything itself and, with no key set, is never called.

What your AI agent sends us

seodraft only receives the arguments your agent sends with each tool call over MCP: a topic name, a draft body, a profile field, and similar. It does not read your conversation with Claude, ChatGPT, or another assistant, their memory, or your files beyond what appears in those specific tool-call arguments.

We do not sell your data or train models on it

seodraft does not sell workspace data. seodraft never calls a model to write an article: drafts come from your agent, over MCP. The only model call in the product is the optional TypeSafe Jev classifier described above, and it answers a narrow question about text you already stored; it does not train on your data.

How long we keep it

Workspace data, which is every category listed above: kept while your workspace exists. After you delete your account, or ask us to delete your data, it is removed within 30 days.

Transactional email log and the delivery events our email provider reports: no job prunes them on a schedule, so they are kept while your workspace exists and are deleted with it, inside the same 30 days.

DataForSEO usage log: the same. It is the spend record for your own account and for the site audits seodraft funded for your workspace, kept while the workspace exists and deleted with it.

Clerk holds your login and email for as long as your Clerk account exists, which deleting your seodraft workspace also ends.

Your controls

Deletion: email chorch@seodraft.app and ask for your workspace to be deleted. It is completed within 30 days, and you get an email confirming it is done.

A copy of your data: email the same address and ask for an export. You get it within 30 days.

Revoking an agent's access: disconnecting seodraft from ChatGPT, Claude, Cursor or any other app revokes that app's access immediately. Your workspace data stays until you ask for it to be deleted.

Credentials: you can remove your DataForSEO login and password, and your GitHub token, yourself from the profile page in the web app. Disconnecting clears what is stored, with no email to us needed.

Contact

Questions about this policy or about your data: chorch@seodraft.app.