We had this idea to create our own framework for building and deploying single purpose AI agents really quickly and easily and would love to get some feedback on it.

It's called the Looped Agent Framework and it is built with 3 core ideas in mind:

- An agent should be a single config file (agent.yaml)

- It should be secure by default so that it doesn't ask for permissions at runtime i.e. all permissions are defined in the configuration. We build with deno to achieve this.

- It is docker native which means you can stand your agent up anywhere that docker runs.

Similar to openclaw/hermes - we also are trying to make it super simple to connect your agent to various messaging platforms or triggers.

I would love to get some feedback from people building agents to understand if this is something this community finds useful - it certainly has been extremely useful for me personally.

The framework itself is open source (Apache 2.0)