Policy Engine
Apply flexible execution rules and policies to each command, device, and environment.
- Role-based policies
- Flexible rule builder
- Access and permission controls
Trusted execution layer for AI & machine control
Execution, approvals, and control for safer AI operations.
Anvil is Membara’s security product brand for controlled AI and machine operations. Membara Commander remains the core tool for identity, policy, approvals, execution, device sessions, audit, and evidence within a separate service boundary.
Identity-bound access
Policy & approval
Traceable evidence
Commander
Operational control
128
Policies
3
Pending approval
8
Device sessions
248
Evidence records
Recent execution
SampleInspect staging health
Identity-bound action
Apply controlled change
Identity-bound action
Production mutation request
Identity-bound action
Execution evidence
Execute safely.
With control and evidence.
Commander trust boundary stays separate from the public website
OIDC identityAuthorityPolicyApprovalEvidence
Anvil principle
“AI should never gain more authority than the identity, policy, approval, and environment allow.”
Anvil carries the security product identity. Identity, enrollment, OIDC, MCP, device grants, execution, audit, evidence, privacy, and service terms remain owned by Commander Core.
Core capabilities
Control execution, not just server connectivity.
Commander binds identity, authority, policy, approvals, execution, and evidence into one auditable control plane.
Approval Flow
Manage approval processes before execution with adaptable workflows.
- Multi-level approval
- Manual or automated approval
- Real-time notifications
Audit Trail
Record complete, tamper-resistant activity history for compliance and investigation.
- Tamper-resistant logs
- Powerful search and filtering
- Export for audit and compliance
Device Sessions
Manage secure connections to servers, IoT devices, and on-premise systems.
- Controlled encrypted access
- Real-time session management
- Support for multiple protocols
Task Execution
Run commands and automated work reliably with clear monitoring.
- Scheduled or on-demand execution
- Retry and error handling
- Real-time status monitoring
Evidence & Logs
Collect execution evidence such as output, files, screenshots, and system logs.
- Centralized evidence storage
- Configurable data retention
- Integration with external systems
How Commander works
From identity to provable execution.
Commander determines who the principal is, what authority it has, which environment it may touch, whether approval is required, and then records the outcome and evidence from the work performed.
01Identity
02Authority
03Policy
04Approval
05Execution
06Evidence
Commander
Operational control
128
Policies
3
Pending approval
8
Device sessions
248
Evidence records
Recent execution
SampleInspect staging health
Identity-bound action
Apply controlled change
Identity-bound action
Production mutation request
Identity-bound action
Authority boundaries
Execution authority is explicit, scoped, and tied to authenticated identity and target context.
Fail-closed control
When identity, policy, approval, target binding, or evidence requirements are not satisfied, execution does not silently continue.
Separate service boundary
Commander keeps its own app, database, OIDC, MCP, operational data, privacy policy, and terms separate from the corporate website.
Membara Anvil Security
Execute with authority. Prove what happened.
Use Anvil to understand the product. Use Commander to authenticate, enroll, connect devices, call MCP tools, approve actions, execute work, and inspect evidence.