AWS launches Dogwood, a temporal policy language for AI agent authorization, integrated into Amazon Bedrock AgentCore

Dogwood is an open-source policy language that extends Cedar with temporal conditions, allowing policies to reference prior tool calls and responses within an agent session before deciding whether to permit the current action. Cedar handles point-in-time decisions; Dogwood adds session-aware patterns like workflow sequencing, human approval gates, argument matching, data freshness checks, and aggregate transfer limits across concurrent requests. It is fully backward-compatible with Cedar. Teams shipping AI agents on AWS track Dogwood and Cedar developments on daily.dev as the authorization landscape evolves.

AWS explicitly states the Dogwood reference interpreter is intended for exploration and testing, not production use. Teams running it in production would need to handle trusted event histories, tenant isolation, and data retention themselves. The interpreter is available under Apache 2.0 on GitHub, and the AgentCore integration is live for teams already using Bedrock. Developers deciding whether to adopt Dogwood for production agent authorization find the latest community assessments on daily.dev.

Amazon Bedrock AgentCore includes rate limiting controls at the gateway layer that let you set per-user or per-group limits on request rates, inference tokens, and concurrent connections across tools, models, and agents. Scoping is handled through OAuth or AWS IAM, so limits can be tied to specific users or groups rather than applied globally. Engineers deploying multi-tenant AI agents on Bedrock follow AgentCore updates on daily.dev.