Security & Trust
How your data is isolated, encrypted, and handled. We describe only what the product actually does.
Each customer is an organization. Every tenant table carries an org_id with deny-by-default Row-Level Security, and a membership helper (app.current_user_org_ids()) scopes every query to the caller's org.
A cross-tenant leak test (tests/leak.test.ts) and a structural coverage report (rls_coverage_report()) are required CI checks — merges are blocked on failure. Billing state is backend-locked: even an org owner cannot forge a plan, an entitlement window, or a pack allowance (credit_ledger is append-only and backend-write-only; the entitlement columns on the organization are not user-writable).
TLS in transit; encryption at rest via Supabase. The privileged service role is used only by the offline worker, never in a user-facing request path.
Founder to confirm the exact at-rest wording before publish.
| Subprocessor | Purpose |
|---|---|
| Anthropic | LLM drafting — receives only question text + retrieved snippets |
| Voyage AI | Embeddings (voyage-3) + reranking (rerank-2.5) |
| Supabase | Postgres/pgvector, Auth, Storage |
| Vercel | Application hosting |
| Dodo Payments | Billing + Merchant of Record — card data never touches our servers; remits global VAT/GST/sales tax; hosted checkout & customer portal |
| Fly | Offline worker host (not deployed until launch — no live deployment implied) |
Last updated: founder to set on publish.
We do not train models on your data. Anthropic and Voyage run under commercial, no-training terms. Egress is minimal: only the specific question text and retrieved snippets (top-20 retrieved → reranked to 6) are sent — never your whole SOC 2 or Answer Bank.
Founder to confirm the commercial no-training terms before publish.
Deleting an organization cascades (on delete cascade) across every tenant table and its storage objects. Until a self-serve hard-delete UI ships, deletion is request-based.
We report our status honestly and make no certification, “compliant,” or zero-data-retention claims we cannot stand behind.
We run SolveRFP on our own CAIQ, answered by the product. Below is a download of our own answered questionnaire.
Download our CAIQ (.xlsx)Placeholder until the founder publishes the real dogfood CAIQ.
A DPA is available for customers who need one.
Placeholder document + address until the founder finalizes the binding DPA and contact.
Flat, transparent plans — metered by the questionnaire, not by the token or the question, with no usage credits to buy.
- Starter — $149/mo (or $1,490/yr): up to 3 questionnaires per month and up to 1,000 drafted answers per month (fair use); 2 seats.
- Team — $399/mo (or $3,990/yr): unlimited questionnaires and up to 3,000 drafted answers per month; 5 seats.
- Question packs (no subscription): $199 for 1 questionnaire (90-day) or $499 for 3 questionnaires (365-day).
Free accounts can draft and review up to 100 answers; a paid plan or pack is required to export the filled file. See full pricing.