Agentrc is an open specification that lets developers package AI agents like container images, with portable OCI artifacts that carry policy declarations and can run on any platform—local, cloud, or Kubernetes—without changes. The specification uses a Dockerfile-like syntax with agent-specific keywords (IDENTITY, CAPABILITY, SOP, POLICY) and enforces governance through Cedar's deny-by-default model, making agent boundaries reviewable and interoperable across cloud providers.
Summaries like this, in your inbox every morning.
Sign up free →What happened
A new open specification called Agentrc has been released, enabling developers to define AI agents using an Agentfile (similar to Dockerfile) that declares the agent's identity, capabilities, tools, and policy requirements in a single declarative format, then compile it into a portable OCI container artifact.
Why it matters
The specification separates agent declaration from execution, allowing the same compiled artifact to run unchanged on local systems, AWS Bedrock, or Kubernetes without modification. Platform operators can review and enforce policies using Cedar's deny-by-default model, addressing governance and security concerns when deploying AI agents across different environments.
What to watch
The project is published as a standards-style repository with specification first and reference tooling second, currently at Working Draft 0.1.0-draft.6. A CLI tool (agentrc) is available for macOS, Linux, and via Homebrew or Go, allowing developers to scaffold, validate, build, and test agents locally before shipping them.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
