Security & Trust

Built for the data you're not allowed to lose

Per-tenant encryption keys you can hold yourself, crypto-shred with legal hold, classification-aware audit, WORM retention, and compliance workflows that ship as single HTML tags. Security is a property of the engine, not a setting someone has to remember.

Envelope Key Management

Your keys. Your tenant. Your call.

Every tenant — every account and every sub-account — owns its own key ring. Data is encrypted under a per-tenant data key that is itself wrapped by a key-encryption key you can keep in your own KMS. Present tenant A's ciphertext as tenant B and it simply does not decrypt.

  • Per-tenant data keys — AES-256-GCM, never stored in the clear, resolved by tenant identity on every read
  • Bring your own KMS — a pluggable key-encryption provider seam shaped for AWS KMS, Azure Key Vault, and GCP KMS; revoke access and the tenant's data becomes cleanly unreadable, fail-closed
  • Rotation without data loss — multiple key versions per ring; old ciphertext still reads, new writes use the new key
  • Crypto-shred with a certificate of destruction — a scheduled, dual-controlled operation, never a single API call
  • Legal hold blocks shred outright — a ring on hold cannot be destroyed, and the refusal is itself audited
  • Flow steps tenant_encrypt / tenant_decrypt let any workflow use the same envelope for its own payloads
 Tenant Key Ring — acme-health
🔑 KEK: customer-held (AWS KMS)Available
↓ wraps
🔐 DEK v3 · active · rotated 12 AugActive
🔐 DEK v2 · retired · still readsRetired
⚖️ Legal hold: ON · shred request refusedAudited
Engine-Level Security

Security built into every layer

FlowClick doesn't bolt security on as an afterthought — it's embedded from the database engine to the web layer. Your data is protected by design, not by configuration.

Zero SQL injection

All queries are parameterized at the engine level — not by developer discipline. There is no way to construct raw SQL through the platform, eliminating injection risk entirely.

Column-level AES-256-GCM

Encrypt individual columns at the application layer — beyond database-level encryption. PII, PHI, and financial fields are encrypted on write and decrypted on read, transparently, and refuse to be queried in the clear.

Pinned, SSRF-safe egress

Every outbound call — HTTP, connectors, webhooks, OAuth — resolves DNS once, validates the address, and pins the socket to the validated IP. Private ranges and loopback are blocked by default.

Multi-factor authentication

Built-in TOTP-based MFA (RFC 6238) with QR code enrollment. One SSI directive renders the complete setup flow — no third-party auth provider required. SAML single sign-on per application.

Content-addressed, verifiable storage

Large objects are stored as deduplicated, checksummed chunks. integrity_verify walks a manifest and proves every chunk still exists and is intact — metadata-only or deep re-hash.

WORM file retention

Immutable object-lock storage with configurable retention periods in compliance or governance modes. Files cannot be modified or deleted until the retention window expires.

Also built in: automatic CSRF validation on every form · role-based access per user, group, and resource · scoped API tokens · auto-hashed passwords · secure session tokens · mandatory TLS · isolated VPC with private routing · auto-provisioned SSL certificates · encrypted Postgres at rest · mTLS with IP whitelisting · continuous protection with point-in-time recovery · cryptographic flow steps (SHA-256, HMAC, AES-GCM, random bytes) so nobody writes their own crypto in a script

Audit & Classification

Everything on the record — and labelled

Three independent audit streams — data, API, and platform-operator — plus a trace spine that records every step of every run. Tables carry a classification so reports, erasure, and access requests know exactly what is sensitive.

  • Row-level data audit — writes always; reads and exports opt-in per table, for when "who looked at this record?" matters
  • Data classification labels — PII, PHI, Financial, Confidential, Internal, Public — drive compliance dashboards and automated workflows
  • Immutable API and operator audit — insert-only; no application role can update or delete an entry
  • Trace spine — one durable run record and one span per step, W3C-shaped trace identity, never a payload or secret in a span
  • Data lineage — a static graph from source field to target column across mappings, connectors, and flows
  • Record protection and retention holds keep specific records from deletion until a policy releases them
 Patients — Table Settings
🏷️ Classification: PHILabelled
🔒 SSN, DOB → AES-256-GCMEncrypted
👁️ Audit reads & exportsOn
📜 1,204 audit entries this weekImmutable
AI Governance

Agents and models, held to the same standard

An AI agent inside FlowClick is a principal like any other: it authenticates, it carries scopes, it is rate-limited, and it leaves a record. Models are versioned nodes with a state machine, not files on a disk.

Permissioned tools

Every agent tool call runs under the calling user's role-based permissions and token scopes — the agent can never do more than the person who invoked it.

Effect budgets & kill switch

Declare per-account caps on classified side effects — emails, SMS, payments, external calls. Each decision to allow, warn, escalate, or refuse is audited. One switch stops all agent activity.

Every run recorded

Chat, job, or in-process — each agent invocation is a durable AgentRun with its principal, tools, and outcome, joined to the same trace as the flow that started it.

Human review before promotion

When an agent proposes a change to an app, connector, mapping, or flow, the platform validates it against real evidence and a person approves it. Nothing promotes on a caller's say-so.

Models under tenant keys

Model artefacts live under the tenant's key ring. Shred the ring and the models become unreadable; a legal hold blocks that shred. Training data never leaves the tenant's instance.

No silent model swaps

Candidates are never auto-promoted to serving. Activation is an explicit, audited transition with a from-state, and champion/challenger routing never changes which version is "active."

How governed agents work →  ·  The model runtime →

Compliance

Compliance-ready from day one

These aren't bolt-ons — they're native platform capabilities with dedicated SSI directives, automated workflows, and full audit trails. Enable what you need per folder with a single setting.

Right to erasure

Automated deletion across all PII/PHI-classified tables matching a data subject identity — hard-delete with full audit logging, extended across documents, files, and model artefacts. One SSI directive renders a public-facing erasure request form.

Supports: GDPR Art. 17 · CCPA · LGPD

Consent management

Purpose-based consent tracking with grant/revoke support — records subject, purpose, timestamp, IP, and user agent. An SSI directive renders a consent banner with per-purpose checkboxes. SMS STOP/START keywords are honoured automatically.

Supports: GDPR Art. 6–7 · ISO 27018 · PIPEDA · LGPD · TCPA

DSAR & data portability

Automated Data Subject Access Requests scan all PII/PHI tables and export a structured file. An SSI directive provides a public-facing data export request form with full audit logging.

Supports: GDPR Art. 15/20 · CCPA 1798.100/110 · PIPEDA

Sub-account shred & legal hold

Multi-tenant hubs can place a whole sub-account on legal hold or shred it cryptographically when a relationship ends — scheduled, dual-controlled, and certified.

Supports: GDPR Art. 17 · SEC 17a-4 retention · litigation hold

HIPAA GDPR SOC 2 Type II SOC 1 Type II SOC 3 PCI DSS ISO 27001 ISO 27017 ISO 27018 SEC Rule 17a-4 SOX HITECH CCPA / CPRA GLBA FERPA PIPEDA LGPD

Business Associate Agreements are in place with our infrastructure providers; Data Processing Agreements and Standard Contractual Clauses are available for international transfers.

Where It Runs

Isolation is an instance, not a checkbox

Data residency and tenant isolation are handled where they can actually be guaranteed — at the infrastructure layer. A tenant lives in exactly one FlowClick instance with its own database, its own object store, and its own configuration, routed by DNS.

  • FlowClick Cloud — managed, scaled, and updated for you
  • Dedicated instance — your own database, bucket, and region for residency or isolation requirements
  • Atoms behind your firewall — a hybrid runtime that executes designated flows and model training entirely on your premises and returns only the result
  • Standalone runtime — deploy exported .hpkg applications to AWS, GCP, Azure, or any OS

More about Atoms →

☁️ FlowClick CloudFastest start
🏛️ Dedicated instance · EU / US / APACResidency
🏢 Atom behind your firewallOn-prem execution
📦 Standalone .hpkg runtimeAny IaaS

Bring your security questionnaire

We will walk your security and compliance team through the key model, the audit streams, and the deployment options — with the engine open.

Start Free Trial Schedule a Demo