Security & architecture

How Henosis keeps customer data safe

This page describes how Henosis REP is built and run today, checked against the live systems on September 7, 2026. Controls that are still ahead are listed in the roadmap at the bottom. The full document is available as a PDF for your IT and compliance reviewers.

The controls in place today

The PDF carries the full document. These are the headline controls.

The database enforces isolation

Every table in the application schema has PostgreSQL row-level security enabled: 115 of 115 tables, measured on production on September 7, 2026. Tenant membership is checked by the database from the verified session token, so a request for another workspace’s data is denied by Postgres, whatever the client sends.

Data is encrypted in transit and at rest

TLS 1.2 or higher on every connection, with HTTP Strict Transport Security. AES-256 at rest for data, backups and storage. Connected-mailbox credentials carry a second layer of AES-256-GCM encryption with the key held outside the database, and the column is locked from the API roles.

Three named people hold production access

Access to the production database and its console is held by three named individuals: the founder and the CTO as owners, and one technical advisor as an administrator. There is no shared login, and every change reaches production through source control and the deployment pipeline.

Changes reach production through the pipeline

Day-to-day work lands on a dev branch that deploys to a development environment. Production is updated only by a pull request from dev, opened and merged by the founder after the pre-promotion audit. Every schema change is a version-controlled migration applied by the pipeline. There is no manual production DDL.

The audit log is append-only

Security-relevant and data-change events are written to an audit log that row-level security keeps append-only: the database admits inserts and reads and denies updates and deletes. Every e-signature action is recorded with actor, IP address, user agent and timestamp. Error monitoring runs through Sentry.

What is still ahead

SOC 2 (Type I, then Type II), an independent penetration test, MFA enrollment in the app, SSO, an enforced Content Security Policy and point-in-time recovery are on the roadmap at the bottom of this page, with a status for each.

Sub-processors

These services process data on our behalf. All are US-based, and customers receive advance notice of material changes under the DPA. A public logo-lookup service receives only a company domain name during brand import.

Sub-processorPurposeLocation
SupabaseDatabase, authentication, storage, server-side functionsUS (AWS us-west-1)
VercelFrontend hosting and CDNGlobal edge, US origin
StripeSubscription billing and payments (PCI DSS Level 1)US
AnthropicAI features: the assistant, briefs and drafting (Claude)US
OpenAIAI features: extraction, note logging and import mappingUS
PerplexityWeb research for lead discovery and enrichmentUS
FirecrawlWeb page fetching for research and brand importUS
ResendTransactional and marketing email deliveryUS
NylasConnected-mailbox send and sync, engagement trackingUS
SentryError monitoring and diagnosticsUS

Security FAQ

The questions vendor security reviews ask most often. If your team uses a formal questionnaire (SIG Lite, CAIQ, or your own), send it and we will complete it directly.

Where is the application hosted and where is data stored?

In a fully managed US cloud. The frontend is served from Vercel’s edge network with a US origin. The database, authentication, storage and server-side functions run on Supabase in AWS us-west-1 (Northern California). Nothing runs on customer-managed infrastructure.

Is the database exposed to the internet?

No. The application reaches the database only through authenticated APIs that enforce row-level security. Direct database connections exist only for the deployment pipeline and the three named individuals.

Is this single-tenant or multi-tenant?

Logical multi-tenant. Isolation is enforced by PostgreSQL row-level security on every table (115 of 115 on September 7, 2026), so a request for another workspace’s data is denied by the database, whatever the client sends.

Is data encrypted in transit and at rest?

Yes. TLS 1.2 or higher in transit, with HTTPS enforced and HSTS preload. AES-256 at rest for data, backups and storage, managed by the providers. Connected-mailbox credentials carry a second layer of AES-256-GCM encryption with a key held outside the database, and only the server-side function that needs a credential can read it.

Do you store payment-card data?

No. Stripe (PCI DSS Level 1) handles all card data. We store billing metadata only: plan, seat count, status, and the card brand, last four digits and expiry as reported by Stripe.

How do users authenticate? Is MFA available?

Email and password through managed authentication, with signed sessions that expire after one hour and refresh tokens that rotate on use. Passwords need at least ten characters with upper case, lower case, a digit and a symbol, enforced on the server, and passwords found in known breach lists are rejected. New accounts confirm their email address before signing in. MFA: the provider supports time-based one-time codes and the capability is switched on, but REP does not yet offer enrollment in the app. It is on the near-term roadmap.

Do you support SSO (SAML/OIDC)?

Not today. SSO is on the enterprise roadmap. Tell us if it is a requirement and we will discuss timing.

Who at Henosis can access customer data?

Three named individuals hold access to the production database and console: the founder and the CTO as owners, and one technical advisor as an administrator. The same three hold source-code access. There is no shared login and no ad-hoc production database access; changes reach production through source control and the deployment pipeline. A role-gated operator path exists for support and provisioning and is recorded in the audit log.

Which AI providers see my data?

Anthropic (Claude) is the primary model and OpenAI is used for selected features. Both are called from the server, never from the browser, and their API terms exclude using the data to train their models. Perplexity and Firecrawl are used for web research in lead discovery and enrichment; they receive search queries and public web content, not your records. AI usage is metered per workspace against a spend cap enforced on the server.

How are changes deployed to production?

Work lands on a dev branch, which deploys to the development environment. Production is updated only by a pull request from dev, opened and merged by the founder after the pre-promotion audit. Every pull request runs a build identical to the production build. Schema changes are version-controlled migrations applied by GitHub Actions; there is no manual production DDL. Branch protection with required reviews and status checks is a planned repository upgrade.

How do you protect against common web vulnerabilities (OWASP Top 10)?

Parameterized queries through the database SDK (no string-built SQL), schema-validated inputs (Zod), React’s default output encoding against cross-site scripting, field-allowlisted CSV and Excel imports (no mass assignment), allowlisted redirect targets on billing and checkout, signature verification on Stripe and Nylas webhooks, rate-limited public signing links with server-issued tokens, and secrets held in provider secret stores. A repository check on September 7, 2026 found no credentials in the source.

Are you SOC 2 certified?

No, not currently. Operational controls are being established now and SOC 2 Type I is targeted after funding. We will share the attestation under NDA when it exists.

Have you had a third-party penetration test?

Not yet. We run internal reviews against the OWASP Top 10 (code review and dependency audit) and a recurring multi-perspective audit with a dedicated tenant-isolation lens; the most recent audits ran on August 21, 24 and 27, 2026. Two internal adversarial reviews in July 2026 produced 64 findings; the critical and high items were fixed and promoted to production, and the open items are listed on the roadmap below. An independent penetration test is planned as part of the SOC 2 program.

Are you GDPR/CCPA compliant? Is a DPA available?

We support GDPR and UK GDPR (lawful bases declared, data-subject rights, US-transfer safeguards) and CCPA/CPRA (no sale or sharing for cross-context advertising). A DPA incorporating EU Standard Contractual Clauses is available on request. HIPAA and GLBA do not apply; we process no PHI or consumer financial-account data.

Who owns customer data and can we get it back?

You own everything you enter; Henosis processes it only to provide the service. You can export your data (CSV or Excel) at any time. On termination there is a 30-day export window, after which the data is deleted and backups age out on their normal cycle.

Are there backups and disaster recovery?

Yes. Automated, encrypted daily physical backups with continuous write-ahead-log archiving, managed by the database platform on its paid plan and kept on a rolling seven-day window. Point-in-time recovery (sub-24-hour RPO) is an available add-on that is not yet enabled; it is on the near-term roadmap.

Is there an audit trail?

Yes. Security-relevant and data-change events go to a tenant-isolated audit log that the database keeps append-only through row-level security (inserts and reads only). Every e-signature action (created, sent, viewed, signed, declined, expired, reminder) is recorded with actor, IP address, user agent and timestamp, and the sent document carries a tamper-evidence hash that is checked at signing.

What is your breach-notification commitment?

We notify affected customers and regulators as required by law, within 72 hours of becoming aware where GDPR applies, with the nature of the incident, the data involved and the remediation steps. A formalized, tested incident-response runbook is being finalized.

Security roadmap

We are an early-stage company. Here is what is not in place yet, in rough priority order, with the status of each item as of September 7, 2026.

  • SOC 2 Type I, then Type II: Type I targeted after funding; controls being established now.
  • Independent third-party penetration test: Part of the SOC 2 program; summary shared under NDA when complete.
  • Multi-factor authentication in the app: Near-term roadmap. The provider capability is on; enrollment in REP is not built yet.
  • SSO (SAML / OIDC): Enterprise roadmap.
  • Branch protection with required reviews and status checks on the production branch: Planned; needs a repository plan upgrade.
  • Full end-to-end test suite as a blocking gate in continuous integration: Planned. The suite runs by hand before a promotion today; a smoke suite runs after every development deployment.
  • Automated dependency and secret scanning, and clearing the open dependency advisories: In progress.
  • Content Security Policy in enforced mode: In progress. The other hardening headers are enforced today; CSP runs in report-only mode.
  • Point-in-time recovery (sub-24-hour RPO) on the production database: Near-term; available as a platform add-on.
  • TLS enforcement on direct database connections: Planned. The application itself never connects directly.
  • Fail-closed rate limiting on the four read-only signing-link endpoints: Fixed on September 7, 2026; ships with the next production promotion. The two signing endpoints already fail closed.
  • Explicit revocation of update and delete grants on the audit tables: Planned, as defense in depth behind the existing policy.
  • Formalized, tested incident-response runbook: In progress.
  • Customer-configurable data-retention controls: Roadmap.

Need the DPA, a SIG/CAIQ response, or more depth?

We answer security questionnaires directly and share additional documentation under NDA on request. The Privacy Policy and Terms of Service cover the legal side.

Self-prepared on September 7, 2026 from a review of the repository, hosting configuration, live endpoints and source code. Accurate to Henosis REP in production as of that date.