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)