Somewhere a company can see and control how AI is used across it: which models are available to whom, what data reaches them, what agents are allowed to do, what it costs, and a record of all of it. The distinction that matters in practice is whether governance covers only the vendor's own product or extends to the AI tools your people already run.

Employees using AI tools the company has not approved, usually on personal accounts, with company data. IBM found 80% of employees do it (IBM, 2025). Blocking pushes it further out of sight. The workable answer is to give people something better that is governed by default, and to route the tools they will not give up.

No, contractually. Data sent to model providers goes by API only, is not retained beyond the time needed to generate a response, and every provider is under a DPA prohibiting use for training. That holds whether you use our keys or your own.

Organization, Space, thread, actor, agent, agent manager, model, sources and cost, for every thread message, every agent action and every Company Brain retrieval, including retrievals made from a third-party AI client over MCP. Searchable and exportable.

Private Space activity is logged like everything else, but it is not shown in the in-app audit browser. Reaching it requires a compliance export under a role granted separately from Admin, and every such export is itself audited. Shared Spaces, including restricted ones, are fully visible to organization admins.

Sensitive data is automatically detected and masked before any prompt reaches a model. It runs pre-model on every call from every Space and every agent, and the rules are administrator-configurable and testable before rollout.

Read or write per agent per connector, not blanket access. An agent acts only where its own permissions and its Space's grants overlap. Each agent has a named manager who is its only editor and appears on every audit line it produces. Revoke in seconds.

Only what each person already had. A request from an outside client is scoped to the user who made it, retrieval is read only so nothing writes back, every retrieval is logged, and admins can disable, restrict or monitor MCP connections organization-wide.

Their model traffic, once routed through the OrgLogic gateway: model controls, per-agent budget caps with hard stops, PII redaction before the model, a per-agent credential you can revoke in one click, and a per-agent audit trail. That scope is model traffic and does not extend to what those clients do outside routed calls.

SOC 2 Type II and ISO 27001, both at the organization level, with a HIPAA BAA available and GDPR compliance. Reports and certificates are available on request for security review.

Yes, all of it. SSO, SCIM, domain verification, VPC and on-premise sit on Enterprise, which carries a 100-seat minimum. Everything else on this page is available from your first user.

Single-tenant VPC and on-premise are available on Enterprise. Production Pilots deploy to a dedicated single-tenant VPC regardless of plan.