Privacy
This page names records the code in this site writes for the public tools, a service request, member sign-in, a waiver, the waitlist, and payments. It also names request logs kept by the host. Other tables exist. This is not a full catalog, and it is not a promise about logs a vendor keeps beyond what is written here.
The public connection
The tools at /api/mcp do not ask you to sign in. They do not place an order and they do not take a payment. A counted call can add one row to mcp_tool_calls. That row is the tool name, a short label for the kind of app that called, a province code from a fixed list (or none), and the time. How to connect is on Connect.
That row does not include your IP address, the raw client name, or the words you sent. A job draft check does not save the draft. If the count cannot be written, the tool still answers and that call is not kept.
Page and click counts
Visits to public agent pages can add to agent_traffic_counts. A row is the page path, a crawler family, the host of the referring site, whether the visit was our own check, which copy of the site served the page, a running total, an hourly count, the time that hour started, and the time the row was last updated. It does not store an IP address, the full browser string, a full referrer link, or an email address.
Starlink page views can add to starlink_daily_counts. A row is the day, a short source tag, which copy of the site served the page, and how many human and bot views and clicks there were. An hourly cap beside those counts stores the source, the kind of count, which copy of the site, a number, and a time.
Referral clicks are also counted in starlink_referral_clicks by a source tag. A row stores a total, an hourly count, the time that hour started, the time the row was last updated, bot hits, a bot hour count, the time that bot hour started, which copy of the site served the page, and whether the click was our own check. These count tables do not store an IP address.
Request limits store a hash and timestamps, not the network address. The service form and the waitlist use service_intake_slots, a hash of the network address. The public connection, the SlipVault pack, and sign-up use slipvault_anon_book_slots, also a hash of the network address. Sign-up also writes a hash built from the account id into that same table. Those rows do not store the address or the account id.
Service requests
The form at /services/request is stored in service_inquiries. A row is what you type:
- the offer and its price label
- business name and town
- your name and email
- phone, a map link or address, and whether a listing is claimed, when you give them
- whether you have a domain: yes, no, or unsure. That answer is not a domain name
- a note, when you write one
- a reference code and the time
When mail is configured, Resend sends a note to the published support address. If a second support inbox is configured and it is a different address, that note goes there too. The address on the form gets a short confirmation. The request stays stored if the email does not send. Resend is not used for member sign-in mail.
If the hidden field on the form is filled, the form is not saved as a request. A durable log, agent_activity_log, stores a hash of the network address and the offer id. If the request table is not available, the full request is stored in that same log instead.
Waiver and waitlist
A signed waiver and a PAR-Q questionnaire can be stored in legal_compliance. That record includes the account id, the business id, the name, the signature, the questionnaire answers, whether a physician warning was shown, a waiver version, the raw IP address, and the raw browser string.
A waitlist row in waitlist_leads stores an email, a product, a lane, a source, and the time it was created.
Payments
A member row can store a Stripe subscription id and a status of active, past due, or canceled. Stripe customer identifiers arrive in the stripe schema through sync, on the synced subscription and charge rows. The Stripe Connect account id is stored as businesses.stripe_account_id. This app writes the subscription id, the status, and that Connect id. It does not write the card number into these tables.
Member sign-in
On the public site, members sign in with an email link or with Google. Other providers may be enabled on test environments. A separate member sign-in uses an email and a password. Supabase keeps the account. The access cookie and the refresh cookie last 7 days. On the email-link and Google path, the sign-in library writes its auth-token cookie after those, and that cookie can last up to about 400 days. That path also sets a PKCE verifier cookie for the handoff, and the library default can keep it up to about 400 days. Choosing a business sets a 30-day selected-business cookie. The email-link and provider sign-in also set a short-lived redirect cookie of 600 seconds so the browser can open the page you asked for. Password sign-in does not set that redirect cookie. The public tools do not read that account. Sign-in mail is sent through Supabase, not through the request-form mail above.
A SlipVault scan on the public page is not saved. The photo is used to build a spreadsheet and is not kept as a file.
Host logs
The site is hosted on Vercel. Vercel keeps request logs. Those logs include the IP address, the browser string, and the path. That is separate from the count tables above, which do not store an IP address.
Contact
Questions about these records: support@phoenix-collective.app, or the form at /services/request.