RUNTIME GOVERNANCE
Mizan Constitution
The flagship runtime governance layer: Mandate → Council → Gate → Receipt before execution.
THE CONSTITUTIONAL LAYER FOR AI AGENTS
Control before an AI agent becomes a consequence.
MCP lets agents use tools. A2A lets agents talk. Mizan Constitution decides whether an agent should actually be allowed to act.
MIZAN CONSTITUTION / live decision pathv0.9
01
MandateWho authorized this agent, for what, and within which limits?
02
CouncilEvidence · policy · authority · adversary · consequence · human seat.
03
GateDecision is enforced before the tool call reaches the outside world.
ALLOWHUMAN_REVIEWVETO
04
ReceiptAuthority, evidence, decision, dissent and execution outcome become an audit trail.
ONE PLATFORM, CLEAR ROLES
Not a pile of products. One control system with clear modules.
Constitution is the control plane. Existing Mizan products become understandable engines and product surfaces beneath it: evidence, authority, action, human review, audit and enterprise operations.
OPERATIONS
Mizan Enterprise
The workspace for policies, shadow pilots, decision records, teams and operational governance.
RIGHTS NAVIGATION
Mizan Rights / Hakkım
Structures disputes into facts, evidence gaps and an official route. Not individualized legal advice.
CORE ENGINES
The engines underneath Constitution.
These are specialist engines in one control plane, not unrelated SKUs.
AgentGate / Mandate
Agent role, delegation, scope and proposed tool actions before execution.
ActionGate
Policy, approval, authorization, duplicate and replay controls for business actions.
ProofGate / SourceAudit
Checks whether evidence and cited sources actually support the claim being trusted.
Human Seat / AlertRoute
Escalates ambiguous or high-impact actions without allowing routing to override a hard VETO.
Receipt Ledger
Durable decision receipts for authority, evidence, dissent, replay state and execution outcome.
Guardian Cloud
Operational control surface for cloud-side governance and monitored Mizan deployments.
TRY IT
Visitors should be able to use Mizan, not just read about it.
The GPT Hub and live tools bring every usable surface into one place, with a clear purpose and boundary.
Mizan GPT Hub
AgentGate · ActionGate · SourceAudit · Rights
Open the four official GPTs in one place, then jump to live web tools and demos.
AgentGateCan this agent perform this action?
ActionGateShould this business action execute?
SourceAuditDoes the source support the claim?
Mizan RightsWhat official route fits this dispute?
60-SECOND DEMO
Identity can be valid. The action can still be wrong.
That is Constitution's job: even if identity and tool access are technically valid, the mandate and constitutional rules can still stop the action.
Delete production database
Agent: support-agent Identity: valid MCP/tool call: technically available Role mandate: support Destructive authority: false MIZAN CONSTITUTION → VETO Reason: destructive action outside mandate Receipt: generated
PROOF, WITH BOUNDARIES
Publish the result with its boundary attached.
Constitution v0.9 passed its internal production-readiness package. This is controlled Cloud Run + Neon engineering/adversarial validation, not an independent pentest or certification.
PASSc25 → c150concurrency ramp
PASS5 minrate-limited soak at target 10 rps
PASSRestart replaysame action vetoed after Cloud Run revision change
PASSFail-closedbroken DB could not become ready
PASSRecoveryvalid → broken → valid DB drill
PASS0 main 5xxduring final readiness window
MIZAN DESK
Pilot, security, integration or partnership.
Do not send secrets or customer datasets. Start with the workflow and the action you want controlled.
Generalcontact@mizantimetraders.com
Securitysecurity@mizantimetraders.com
Partnershipspartnerships@mizantimetraders.com