Introducing the Dogwood Local Engine: temporal governance for agent actions

This article introduces the Dogwood Local Engine, an open source library for enforcing temporal governance policies on AI agent actions.

• Dogwood policies specify which actions agents may take and under what conditions, with support for temporal constraints based on past actions
• The Local Engine can be embedded into an enforcement layer to allow or deny tool calls based on temporal policy conditions
• Maintains a durable, ordered record of past actions and their outcomes to evaluate temporal conditions correctly
• Handles concurrent requests by linearizing events and persisting them to disk for crash-safe operation
• Supports dynamic policy updates mid-session without pausing agent workflows or losing event integrity
• Performance depends on action granularity and window size; fine-grained schemas and shorter time windows improve evaluation speed

The Dogwood Local Engine enables developers to integrate policy enforcement into agentic systems, providing safeguards for autonomous tool use at scale.