Two years ago, I was stuck. Working a job that didn't fit, living through some rough times in Brazil, then moving to Italy after COVID looking for a fresh start. I needed change - not just geographically, but mentally.
I went deep on personal development, habits, mindset. Then AI happened. I saw the opportunity and thought: "I can actually do this."
But working with AI agents was... chaotic. Context limits. Lost state. Switching between conversations. Repeating yourself 50 times. I kept thinking: "There has to be a better way."
So I built CervellaSwarm. 260+ work sessions later, it's real.
*What it is:*
- 16 specialized AI agents (1 orchestrator "Regina" + 3 quality/security/ops guardians + 12 workers)
- MCP-native (Model Context Protocol) - agents communicate like a real team
- Privacy-first: your code stays local, no cloud sync required
- CLI + MCP server ready to plug into Claude Desktop
*What makes it different:*
- Agents have persistent memory across sessions (SNCP system)
- They coordinate - backend talks to frontend, marketing validates UX before implementation
- Built by someone who actually uses it daily for 3 production projects
Hey HN! I'm Rafa, the founder of CervellaSwarm.
Two years ago, I was stuck. Working a job that didn't fit, living through some rough times in Brazil, then moving to Italy after COVID looking for a fresh start. I needed change - not just geographically, but mentally.
I went deep on personal development, habits, mindset. Then AI happened. I saw the opportunity and thought: "I can actually do this."
But working with AI agents was... chaotic. Context limits. Lost state. Switching between conversations. Repeating yourself 50 times. I kept thinking: "There has to be a better way."
So I built CervellaSwarm. 260+ work sessions later, it's real.
*What it is:*
- 16 specialized AI agents (1 orchestrator "Regina" + 3 quality/security/ops guardians + 12 workers) - MCP-native (Model Context Protocol) - agents communicate like a real team - Privacy-first: your code stays local, no cloud sync required - CLI + MCP server ready to plug into Claude Desktop
*What makes it different:*
- Agents have persistent memory across sessions (SNCP system) - They coordinate - backend talks to frontend, marketing validates UX before implementation - Built by someone who actually uses it daily for 3 production projects
*Try it:*
npx cervellaswarm init npx cervellaswarm task "add login page"
Apache-2.0 licensed. Uses your own Anthropic API key.
*What I need from you:*
- Does this solve a real problem for you? - What would make you actually use it vs regular Claude? - Any dealbreakers in the architecture?
GitHub: https://github.com/rafapra3008/CervellaSwarm npm: https://www.npmjs.com/package/cervellaswarm