SOC 2 Overview

A factual map of which SOC 2 Trust Services Criteria this platform already has real, working controls for, and which ones still need work, decisions, or a third party before a formal audit would be realistic.

This platform is not SOC 2 certified, and this page is not an attestation. SOC 2 is not something a codebase "has", it is an audit opinion issued by an independent CPA firm after examining controls over a period of months. What follows is an honest inventory of the technical controls that already exist, organized the way an auditor would look at them, so you can see exactly how much of the groundwork is done and what remains before a Type I (design) or Type II (operating effectiveness over time) audit would be worth commissioning.

1. Security

Control areaCurrent state
Multi-factor authenticationEnforced on every password sign-in, not optional. A 6-digit code is emailed and must be verified before a session is created. Note for an auditor: this is an email-delivered one-time code, not a TOTP authenticator app, which is a meaningful distinction to record accurately in any control narrative.
Password storageBcrypt with 12 rounds.
API key lifecycleA hard maximum expiry of 90 days is enforced server-side on every key, regardless of what a caller requests. Rotation is supported with a 24-hour grace period so a live integration does not break the moment a key is rotated.
Rate limiting / brute-force protectionA global request-rate limiter, plus specific limits on one-time-code sending and verification attempts. There is currently no account-lockout mechanism separate from rate limiting.
Access control modelEnforcement exists and is real, and is implemented as two mechanisms by design rather than one unified system: the API enforces per-API-key scopes for machine callers (a key can be limited to only the ledger, only leads, and so on), while the customer portal enforces role checks for signed-in human users. Those are different authentication models, not the same check written twice, so a single merged layer is not the goal. Within the portal side, though, the role checks themselves used to be re-declared inline in multiple files (team, billing, integrations); they now come from one shared permissions module (`apps/portal/src/lib/access-control.ts`), removing the risk of two copies of the same rule drifting apart.
Webhook authenticityBoth incoming Stripe webhooks and outgoing platform webhooks are cryptographically signed and verified.
Tenant isolationEvery account's data is scoped by account ID at the query level, and this has been specifically tested for cross-account access attempts, including via API key manipulation.

2. Availability

Control areaCurrent state
Uptime monitoringA real, automated health check runs every 5 minutes and is stored, not a manually maintained status page. See the live figures on the Status page.
BackupsAutomated daily database backups with 14-day retention exist and are exercisable via a restore tool. Backups are compressed and encrypted at the file level (AES-256-GCM) before being written to disk, and an optional automatic upload to off-site cloud storage is now supported. One gap worth flagging for an auditor: off-site upload is opt-in via configuration, so a deployment that has not set it will still keep backups on local storage only.
Queue isolationEach background job type (intake, qualification, scheduling, payment, and so on) runs on its own isolated queue, so a failure or slowdown in one does not block the others.
Load testingA load-testing tool exists and has been run at up to 100 simulated concurrent users against the lead pipeline.
See Status for the live uptime numbers, and Support for the current escalation-tier caveats.

3. Processing Integrity

Control areaCurrent state
Financial ledger integrityEvery ledger entry is chained by a cryptographic hash of the entry before it, and the full chain is automatically re-verified every hour. This detects both an entry edited in place and an entry deleted, inserted, or reordered, full detail in the Settlement System page.
Concurrency safetySimultaneous writes to the same account's ledger are serialized using a database-level lock, verified under concurrent load, so two payments arriving at the same instant cannot corrupt each other's balance.
Idempotent retriesSensitive write operations (like retrying a settlement) support an idempotency key, so a repeated request from a network timeout cannot be applied twice.

See Settlement System for the full detail behind both of these.

4. Confidentiality

Control areaCurrent state
Encryption in transitStrict-Transport-Security is set in the application to enforce HTTPS for the browser going forward. The actual TLS certificate and cipher configuration is handled at the infrastructure layer (the web server in front of the application), which is outside what this codebase can attest to directly, that piece should be confirmed separately as part of any audit.
Encryption at rest for personal dataSpecific fields (a lead's email, phone number, and message) are individually encrypted with AES-256-GCM before being written to the database, not relying solely on whole-disk encryption. Two honest limitations: this currently covers those specific fields on the lead record, not every personal-data field in the system; and the encryption key is a single environment-level secret rather than a managed key with rotation.
Audit loggingEvery sensitive action (creating a key, changing a role, revoking access) is written to an audit log capturing who did it, from where, and when. This is append-only by convention today, meaning no application code ever updates or deletes a log entry, but this is not yet enforced at the database level with a hard constraint, and there is no defined retention period in code yet.

5. Privacy

Control areaCurrent state
Data export (subject access)A real, working export covering a customer's leads, ledger history, workflows, and audit trail, available as a self-service download.
Data deletionDeleting an individual lead is a real, working operation. Deleting an entire account cascades and removes its related records. A defined, documented process and turnaround time for a customer's formal deletion request has not yet been written, that is a policy step, not a technical one, and is covered in the Data Processing Agreement page.
Sub-processor disclosureA current list of every third party that touches customer data is published, and is kept matched to the real infrastructure rather than a generic template.

6. What stands between here and a formal audit

None of the items below are contradicted by the platform, they are the honest remaining gap between a genuinely solid set of technical controls and something an outside auditor could sign off on:

  • A written information security policy exists now (see the Information Security Policy), covering access control, data classification, and incident response, with a self-assigned interim owner and a self-conducted initial review already on record. What is still missing is a permanent owner named by the client and a formal external sign-off, not the policy or the review itself.
  • A defined data retention and deletion turnaround commitment, in writing, this is called out explicitly as still open in the Data Processing Agreement.
  • A formal, scheduled access-control review process. Portal role checks that used to be duplicated ad hoc across the team, billing, and integrations pages now come from a single shared permissions module, so there is one place to audit rather than several that could silently drift, and the engineering team now self-reviews it quarterly per the Information Security Policy. An independent, externally audited version of that review is the piece still open.
  • Off-site backup storage now exists as a supported, opt-in upload path (see Availability above), but is not yet turned on by default for every deployment. That is a configuration step, not a missing capability.
  • A decision on whether TLS/certificate management should also be brought under this platform's own documented control, or formally delegated to the hosting provider with that delegation recorded.
  • Selecting and engaging an auditor once the above is in place, starting with a Type I assessment before a Type II is realistic.

Was this page helpful?