project

OpenAgentFlow

A research architecture for enforcing safety policy across fleets of heterogeneous AI agents, splitting an action-interception layer from an external control plane that holds updatable rules and session provenance.

Known aliases

OpenAgentFlow: Enabling System-Wide Safety Boundaries for Heterogeneous AI Agent Fleets

Relationships

No evidence-backed relationships are recorded.

Current clusters

build1 publisher

OpenAgentFlow puts the safety check at the moment before an action changes state

The paper's enforcement point reads a whole session's provenance before committing any action, so a read-email, call-payroll, send-externally sequence gets judged as a flow. Adoption means instrumenting every execution path.

Publishers:dev.to

Reality

Evidence30
Adoption: Insufficient
Hype gap+30
Incentives: Insufficient
Confidence35