AWS Open-Sources Dogwood, Extending Cedar to Govern Sequences of Agent Tool Calls

7.6 relevance

New AWS open-source policy language for agent tool call sequences is highly relevant and novel.

AI/ML infoq.com

Answer the board on AI readiness

Arrive with evidence, not opinion. See how your foundation benchmarks against your peers, get a clear read, and leave with a ranked place to start.

Why It Matters

For platform engineers building multi-agent systems, Dogwood addresses a critical gap: stateless policy engines cannot enforce workflow-level constraints like 'stop after $5k total' or 'require approval before external API calls,' which are essential for safe agent orchestration.

Key Takeaways

Evaluate whether your agent workflows need temporal constraints (e.g., cumulative limits, sequential gates) and adopt Dogwood for those policies while keeping Cedar for stateless, formally verifiable access control.

AWS Open-Sources Dogwood, Extending Cedar to Govern Sequences of Agent Tool Calls

Summary

AWS open-sourced Dogwood, a policy language for agent tool calls that extends Cedar with temporal conditions capable of inspecting an agent's event history. Unlike Cedar's stateless, single-request evaluation, Dogwood supports operators like `formerly`, `count_within`, and `sum_within` to enforce constraints across sequences—such as cumulative spending limits or approval gates before sensitive actions. The language ships under Apache 2.0, is supported by AgentCore Policy, and is built as a separate language to preserve Cedar's formal verification capabilities, which temporal conditions forfeit.

Author

Steef-Jan Wiggers