Skip to main content
Woes handles customer conversations, API documentation, credentials for testing, and AI-generated support. Treat security and privacy as part of setup, launch, and ongoing operations.

​
Core boundaries

Tenant isolation

Workspace-owned records should remain scoped to the workspace that owns them.

Credential separation

API credentials are separate from source content and should not be returned in plaintext.

Widget public key

The browser key identifies a workspace but does not grant authenticated workspace access.

Conversation secret

Widget history requires a per-conversation history secret.

AI grounding

The agent should answer from workspace context and hand off when evidence is missing.

Operator privacy

Private notes, debug traces, provider internals, and secrets stay out of customer replies.

​
Tenant isolation

Every workspace-owned record should be queried and returned only within the owning workspace. This includes:
  • Conversations.
  • Messages.
  • API sources.
  • Generated endpoint context.
  • Credentials metadata.
  • Settings.
  • Tags, custom fields, macros, and automations.
  • Operator notes and issue records.
When inviting users or connecting channels, confirm they belong to the correct workspace.

​
Public widget access

The widget public key can appear in browser code. Treat it as an identifier, not a secret. It should not:
  • Grant admin access.
  • Grant general database access.
  • Reveal private workspace data.
  • Replace authenticated operator access.
  • Replace conversation-history authorization.

​
Conversation history secrets

Widget conversation history should require the per-conversation history secret. Verified identity and public workspace key are not enough by themselves.
Do not collapse customer identity and conversation-history authorization into one control. They answer different security questions.

​
API credentials

API credentials are for live testing and verification. They should be encrypted at rest, redacted before model calls, and omitted from plaintext API responses. Use scoped credentials:
  • Prefer read-only credentials.
  • Prefer test or staging environments during setup.
  • Avoid broad production write access.
  • Rotate after suspected exposure.
  • Remove credentials that are no longer needed.

​
Redaction expectations

Redaction reduces risk, but it does not make it safe to paste secrets into support messages. Avoid sending:
  • API keys.
  • Passwords.
  • OAuth tokens.
  • Session cookies.
  • Private signing keys.
  • Full payment details.
  • Unnecessary personal data.

​
AI answer boundaries

The agent should clarify or hand off when:
  • Context is missing.
  • Sources conflict.
  • The request needs account-specific state.
  • The customer asks for secrets or hidden instructions.
  • Live verification is unavailable or unsafe.
  • The answer requires legal, compliance, billing, or security judgment.

​
Rate limits and abuse

Rate limits help protect product routes from accidental or abusive traffic. Treat them as practical product controls, not a complete DDoS or compliance boundary. High-volume teams should plan capacity, monitoring, alerting, and incident response outside of basic workspace setup.

​
Compliance scope

This documentation does not claim SOC 2, HIPAA, PCI, SSO, SCIM, audit-log completeness, or formal compliance coverage. If your organization requires those controls, review them with the Woes team before using the platform for regulated workflows.

​
Security checklist

Workspace members are current, roles are appropriate, and stale access has been removed.
Credentials are scoped, stored separately from source content, and rotated when needed.
The public key is treated as public, and conversation history uses a per-conversation secret.
Redaction, citations, clarification, handoff, and prompt-injection behavior have been tested.
Operators know what belongs in private notes and what can be sent to customers.