Welcome — what this Discussions board is for

[dep0we]

Hi — Dan here, the maintainer.

atomic-agents-stack is a framework for AI agents whose entire state lives in a folder of Markdown files on your disk: persona, memory, journal, audit log, tools, cost guardrails. No external database. Same agent definition runs from cron, a Claude Code skill, a ChatGPT skill, or embedded in Python.

Where things are right now:

• Alpha, pre-1.0, single maintainer
• 8 of the planned backend Protocols shipped (Memory, LLM, Judge, Lock, Log, AgentProfile, ToolRegistry, Mandate)
• 23 locked spec docs + a working Python reference implementation
• Caldwell sample includes 5 days of real run logs + cost dashboard

What to use this board for

• General — "Is this for me?" / first impressions / general questions
• Ideas — proposals for new capabilities, especially ones that need discussion before becoming an issue
• Q&A — usage questions, "how do I…", "what does the spec mean by…"
• Show and tell — what you're building, even if pre-implementation
• Announcements — releases + bigger direction posts (mostly maintainer-posted)

When to file an Issue instead of post here

• You hit a bug with a clear reproduction → bug issue template
• You have a concrete feature ask with a clear shape → feature issue template
• You're not sure which → post here, we'll figure it out together

Introduce yourself if you want

If you landed here, I'd love to know: what brought you to this project, and what problem are you trying to solve? No pressure — lurking is fine too.