Skip to main content

Homebrew for Personal AI Infrastructure

The 10-star vision for ModusBrain’s integration system. Ship Approach B (v0.7.0), build toward this over subsequent releases.

The Vision

ModusBrain becomes a personal infrastructure operating system where every signal in your life flows through the brain automatically. Integrations are senses (data inputs) and reflexes (automated responses to patterns). Users subscribe to the creator’s actual operating system, then customize it.
The user feels: “My brain is alive. It’s watching everything I care about, and it’s getting smarter every day. I didn’t have to write any code. I just said yes when the agent asked.”

Architecture: Senses & Reflexes

Recipe Format (YAML frontmatter + markdown body)

Dependency Graph

Recipes declare requires in frontmatter. The CLI resolves dependencies before setup. If voice-to-brain requires credential-gateway, the agent sets up credential-gateway first.

Health Dashboard

modusbrain integrations doctor runs health_checks from every configured recipe:

Sense Analytics

modusbrain integrations stats aggregates heartbeat data:

Reflex Rules Engine (future)

Reflexes are recipes that trigger on brain state changes:

Roadmap

Key Design Decisions

  1. ModusBrain is deterministic infrastructure. Cross-sense correlation, pattern detection, and intelligent responses are the agent’s job (OpenClaw/Hermes). ModusBrain provides the plumbing.
  2. Agents ARE the runtime. No npm packages, Docker images, or deterministic scripts. The recipe markdown IS the installer. The agent reads it and does the work.
  3. Very opinionated defaults. Ship the creator’s exact production setup as the default. Users customize from there. Unknown callers get screened. Quiet hours are enforced. Brain-first lookup happens on every call.
  4. Agent-readable outputs. All CLI output must be parseable by agents (—json flag). Migration files include agent instructions. The agent is the primary consumer, not the human.