According to The Register, AWS has released Dogwood Local Engine (DLE), an open source Rust library that can be embedded into an AI agent's harness or gateway. DLE evaluates each tool call against policies written in Dogwood, the open source governance language AWS published in August, and returns an allow or deny verdict. The harness itself is responsible for enforcing that verdict by intercepting every tool call and running the tool only when allowed.
DLE's distinguishing feature is temporal awareness. It logs tool-call events over time and persists each entry to disk before evaluating policy, so the engine can retain state across crashes or restarts. AWS gives the example of a coding agent's Git push: a policy can require that the most recent test run passed and that the pass occurred within the past 15 minutes; otherwise the push is denied.
AWS says the added delay is minimal. In tests of sessions from five minutes to 12 hours, DLE evaluation took about 20 microseconds with a 15-minute window at the 12-hour mark, and about six milliseconds with a 24-hour window. A lock allows only one event submission at a time through policy evaluation, but The Register notes that AWS did not clarify how it would handle concurrent submissions where one would pass and one would fail; AWS did not respond to that question before publication.