Recommended · Live evaluation
External Evaluation Workspace
An active Evaluation grant now carries two separate controls: an issued ZAXAM launch for workspace entry and an issued Evaluation API key for execution authority.
Flagship project · Governed evaluation
Maestro
An Agentic Operations & Governance Control Plane that makes authority, quota, execution policy and audit evidence visible around AI-assisted workflows.
IdentityAuthorityEntitlementsQuotaRuntime CapacityGovernanceIntegrationAudit
Choose your path
See the concept — or test the governed runtime
The interactive demo explains the product without credentials. The External Evaluation Workspace is the real sandbox evaluation surface for approved evaluators.
No credentials · Deterministic
Interactive Maestro demo
Explore synthetic tasks, governance decisions, human-validation gates and audit evidence without login, key or backend runtime.
Governed External Evaluation
From issued authority to durable proof
The Workspace remains on the dedicated Maestro Evaluation origin. ZAXAM-TECH hosts the controlled launch wrapper but never becomes the credential or execution authority.
1 · LaunchUse the supervisor-issued one-time ZAXAM workspace launch.
2 · ConnectUse the separately issued Evaluation API key.
3 · ExecuteA fresh admitted execution consumes +1 quota.
4 · ProveReview safe durable evidence and audit outcome.
ZAXAM-only framed launchSeparate API execution authorityPostgreSQL authorityAdmitted-execution quotaRevocable keyDurable evidenceProduction disabled
Processual Maestro · External Evaluation WorkspaceWorkspace entry is fail-closed until the supervisor issues a one-time launch.
SANDBOX · CONTROLLED ENTRY
Security boundary: the launch wrapper runs on ZAXAM-TECH and does not store the launch ticket or Evaluation API key. The actual workspace executes on the dedicated Maestro Evaluation origin. Direct Render navigation is not an authorized workspace path; task execution still requires the separately issued Evaluation API key. Production execution remains disabled.
Before you connect
You need an active Evaluation grant, a current one-time workspace launch from the supervisor, and the separately issued Evaluation API key. A commercial subscription is not required.
What consumes quota
A new admitted execution consumes one unit. Status reads, durable replay and idempotency conflict consume zero additional units.
What remains unavailable
Direct workspace entry without launch authority, production execution, and unsealed tasks or bindings remain blocked. Revocation ends authority while historical evidence remains reviewable.
Why Maestro
Govern execution, not just prompts
Organizations can prototype agents quickly; operational adoption is harder. Maestro separates authorization and commercial rights from execution and connects each runtime decision to identity, entitlements, quota, capacity, governance, integration and audit.
Authority before admission
Tasks, bindings, quota and environment boundaries are sealed into explicit authority rather than inferred from browser state.
Fail-closed runtime
Sensitive operations stop when required authority, readiness or evidence is missing instead of silently assuming success.
Enterprise integration
Adapter-oriented contracts support controlled sandbox reads and governed draft or approval-gated operations.
Evidence as an output
Durable evidence and audit trails are treated as first-class execution results rather than secondary logs.
Readiness
Evaluation-ready, not production-certified
Readiness note: the External Evaluation path is a sandbox qualification surface. Production rollout, customer-specific integration, commercial readiness and compliance certification remain separate gates.