Trust Center
Odin handles the most confidential documents a firm owns. This page is the whole posture in two honest lists: what is in place today, and what is committed — never a claim in between.
Why firms trust Odin with deal data
- No model training on customer data.Odin's AI calls the Anthropic API, which does not train on API inputs.
- Compartmented Mode. A deal can be walled off so firm-wide AI features never read it.
- Per-deal scoping. Admins control which members can reach which deals — at invite time, org-wide, or deal by deal.
- Append-only audit trail. Sensitive actions are recorded with actor, IP, and user agent.
In place today
| Control | What it means |
|---|---|
| Transport encryption | TLS 1.2+ with HSTS on every customer-facing endpoint. |
| Encryption at rest | Database and object storage encrypted at rest by our infrastructure providers (Supabase, Vercel), AES-256 by default. |
| Access control | Role-based permissions checked server-side on every API route, with Postgres row-level security as a second line of defense. Optional per-deal scoping restricts which members reach a given deal. |
| Multi-factor authentication | Available on every account via Clerk; workspace admins can require it for their org. |
| Audit trail | Append-only audit log of sensitive actions with actor, IP address, and user agent — never updated or deleted by application code. |
| Rate limiting | Per-org request rate limits on the API (Upstash). |
| Change management | Six required CI checks on every change (typecheck, lint, tests, build, security guards, dependency audit); production deploys only from the protected main branch; commit-time secret scanning; Dependabot CVE monitoring. |
| AI spend & usage controls | Per-org metering on every AI call, monthly caps with automated spike alerts, and per-user ceilings. |
Committed — not yet in place
Listed separately on purpose: none of the items below appear anywhere on this site as a current capability, and each moves up only when it is real.
- SOC 2 Type 1. We engage an auditor as customer demand warrants, and claim it only when the audit begins.
- Web application firewall. Platform WAF evaluation is scheduled on the security roadmap.
- Static analysis (CodeQL / GHAS). On the security roadmap behind the CI gates already in place.
- Third-party penetration test. Ahead of the SOC 2 engagement.
- MFA enforced by default. Org-required enrollment with step-up verification on sensitive admin actions.
Continuity & incident readiness
- Incident response. A written IR plan defines severity levels, roles, and customer-notification commitments; it is exercised by walkthrough. Security contact below.
- Backups & recovery. Daily automated database backups via Supabase, with documented recovery objectives in our BC/DR plan.
- Policy set. An information-security policy set and a maintained risk register govern engineering practice; public-safe summaries are available to customers under NDA on request.
Vendors & data flow
10 sub-processors touch customer data, each listed with role and region on the sub-processor register. The full picture — what data lives where, and why — is on the Security page, alongside our DPA, Privacy Policy, and Terms.
Report a security issue
Email security@txnhub.app. We acknowledge within one business day. We do not yet run a formal bounty program.
Last updated: 2026-08-11