Skip to main content
How It WorksSandboxPricing
← All features

MULTI-AGENT GOVERNANCE

Multi-Agent Arbitration

When Agent A delegates to Agent B, RANKIGI verifies authority from the human owner and logs the entire delegation chain.

How it works

In multi-agent systems, agents routinely delegate tasks to other agents. Without governance, these delegations are invisible. RANKIGI’s arbitration layer intercepts every agent-to-agent delegation and verifies that the target agent has authority from the initiating agent’s human owner before the action executes. The verification decision, permitted scope, and trust score are all logged as a single governance event on both agents’ audit trails. Organizations can configure arbitration policies — requiring explicit human approval, auto-approving trusted agents, or setting maximum delegation depth.

Technical details

arbitration_hash = SHA-256(canonical_json({ initiator_passport_id, target_passport_id, requested_actions, approved_actions, resolved_at, human_owner_id })). Logged to both agents’ event chains. Max delegation depth configurable per policy.

Compliance mapping

EU AI Act Article 14SOC 2 CC6.1NIST AI RMF

Use cases

  • Orchestrator agents delegating research tasks
  • Payment agents requiring human approval for high-value transactions
  • CI/CD agents delegating deployment approvals

Related features

FederationHuman AccountabilityPolicy Marketplace
Start governing your agents →Read the standard →