dep0we/atomic-agents-stack
dep0we
Python1updated 4mo ago
Python1updated 4mo agoAI agents that live in your folder, not someone else's database. Vault-native, MIT-licensed. 21 spec docs + Python reference implementation.
ClaudeCodexCursorMCP
agentsai-agentsanthropicclaudeframeworkllmmarkdownmcpopenaipython
README
atomic-agents-stack
AI agents that live in your folder, not someone else's database.
Vault-native, MIT-licensed, Markdown-source-of-truth.
<p align="center"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/dep0we/atomic-agents-stack/main/docs/assets/atomic-agents-hero-dark.svg"> <img src="https://raw.githubusercontent.com/dep0we/atomic-agents-stack/main/docs/assets/atomic-agents-hero.svg" alt="Atomic Agents at a glance: an agent is a folder of Markdown files (persona, tools, memory, wiki, journal, log); the runtime is stateless and wrapped in cost guardrails; every run writes a JSONL audit line, a typed memory note, and a journal entry. Same agent definition runs from cron, launchd, a Claude Code skill, or embedded in Python." width="100%"> </picture> </p>New here? Read the visual guide — a 30-minute plain-English walkthrough of what an atomic agent is, how it runs, the safety layer, and the protocol-pattern scaling story.
Why this exists
Your AI agent's persona, memory, and audit trail live in someone else's database. The persona is locked inside Letta's hosted memory blocks. The memory is in Mem0's vector store. The audit trail is in LangSmith. The cos