The goal is not “another chatbot”, but an agent that can:

decompose goals

plan actions

use tools

reflect on failures

and improve its behavior over time

Think long-running agents, not prompt → response.

What Orcbot does today

Plugin-based architecture (skills are first-class)

CLI + messaging integrations (Telegram / WhatsApp)

Autonomous execution loops

Error recovery & retry logic

TypeScript / Node.js, MIT licensed

Repo: https://github.com/fredabila/orcbot

Why I’m posting

The project has reached the point where architecture decisions matter more than code volume, and I’d love feedback and contributors who care about:

agent planning & orchestration

memory systems (episodic / vector)

tool-using agents

self-correction and evaluation loops

multi-agent coordination

I’m especially interested in people who’ve built:

production bots

autonomous systems

developer tools

or have opinions on how agents should fail and recover

What I’m explicitly not claiming

This is not AGI

This is not magic

This is an evolving experiment in practical autonomy

How to get involved

Technical feedback in this thread is very welcome

Issues and PRs on GitHub

If you’re curious but unsure where to start, open a discussion — I’m happy to guide

I’m posting this mainly to learn from the HN community and see where this direction breaks or shines.

Thanks for reading.