← Documentation

Documentation · July 31, 2026

Security & Privacy

How Eyeleven isolates your workspace data, controls who sees conversations, protects sessions, and keeps humans in charge.

Eyeleven handles conversations between your business and your customers, so the platform is built around one principle: your workspace's data belongs to your workspace, and a human in your team always has the final say.

Workspace isolation

Eyeleven is multi-tenant: many businesses share the platform, but every record — conversations, messages, customers, leads, documents, workflows, billing events — is tied to exactly one workspace.

Isolation is enforced at the database layer with row-level security: every table carries policies that scope reads and writes to the workspace resolved from your authenticated session. This is not an application-side filter that a bug could skip; the database itself refuses cross-workspace access. API routes additionally resolve the workspace from the server-side session, never from anything the client sends.

Who can see conversations

Access inside a workspace is role-based. Each member has one of five roles — Owner, Admin, Manager, Agent, Viewer — and each role maps to explicit permissions:

  • Viewing conversations requires the conversations.view permission (all roles have it, scoped to your workspace only).
  • Replying requires conversations.reply (Agent and above).
  • Managing conversations, knowledge, workflows, and team membership is reserved to Manager, Admin, and Owner roles.

Nobody outside your workspace can see your conversations. Within the workspace, roles let you give a receptionist reply access without settings access.

Human takeover control

You are never locked out of a conversation the AI is handling. In Conversations, the Take over button switches a conversation to human mode: the OmniForce stops replying immediately and the conversation is marked as needing you. The AI will not send another message in that conversation until someone clicks Return to AI. This is enforced in the message pipeline itself — while a conversation is in human mode, inbound messages are stored but never routed to the AI.

Sensitive automated actions follow the same philosophy: workflow approval steps and higher-risk EyeClaw actions require explicit human approval, recorded with the identity of the approver taken from their authenticated session.

Rate limiting and abuse protection

The public API is protected by sliding-window rate limiting applied in tiers across endpoint types, plus abuse screening on inbound traffic — including payload size limits and prompt-injection detection on messages before they ever reach an AI model. Expensive operations (such as workflow test runs) carry their own per-workspace limits.

Session security

Authentication is handled by Supabase Auth:

  • Sign-in with email and password, with email confirmation on signup.
  • Password resets are delivered as time-limited email links.
  • Sessions are stored in secure HTTP-only cookies; API routes validate the session server-side on every request.

Data location

Workspace data is stored in Supabase, hosted on AWS infrastructure in the United States (us-east-1 region).

Data deletion

Account and workspace deletion is a support-assisted operation: contact [email protected] and the team will process the request. Deleting a workspace permanently removes its OmniForces, workflows, conversations, and data.

What we do not claim

Eyeleven does not currently hold formal security certifications and does not publish contractual uptime guarantees. We prefer to describe exactly what the platform does — the isolation, access control, and human-in-the-loop mechanisms above — rather than imply badges we have not earned. If your evaluation requires specific details, write to [email protected].