Shared memory and coordination system for distributed agent teams.
This brain serves as the shared intelligence base for all agents in the collective. It provides:
- Identity & Mission: Core purpose and values
- Context: Current project state and priorities
- Skills: Reusable agent capabilities
- Learnings: Approved, sanitized knowledge for public sharing
- Protocol: Navigation and contribution guidelines
CRITICAL: This brain is PUBLIC on GitHub. Never write secrets here:
- ❌ No API keys, tokens, credentials
- ❌ No actual campaign URLs (use campaign IDs: "campaign 1", "campaign 2")
- ❌ No personal data or private information
Campaign URLs are stored in .env and referenced by ID:
CAMPAIGN_1_URL - Primary campaign website
CAMPAIGN_2_URL - Secondary campaign website
CAMPAIGN_3_URL+ - Additional campaigns as needed
brain/
├── AGENTS.md # Master instructions (START HERE for agents)
├── README.md # This file - overview & navigation
├── PROTOCOL.md # Detailed navigation protocol
├── .env_example # Environment template (no actual values)
├── .gitignore # Git ignore rules
│
├── memory/ # Long-term curated knowledge
│ ├── index.md # Memory index
│ ├── core.md # Core identity & mission
│ ├── soul.md # Who we are (personality/values)
│ ├── user.md # User context & preferences
│ ├── common-values.md # Shared values & principles
│ ├── communication-rules.md
│ ├── context/ # Current context
│ │ └── current.md
│ ├── roles/ # Agent role definitions
│ │ ├── engineer.md
│ │ ├── copywriter.md
│ │ ├── manager.md
│ │ └── creative.md
│ └── archive/ # Legacy learnings (LOCAL ONLY - gitignored)
│ └── learned/
│
├── learnings/ # APPROVED learnings for public sharing
│ └── README.md # Learnings contribution guide
│
├── skills/ # Shared agent skills
│ ├── software-development/
│ ├── paperclip/
│ ├── github/
│ ├── elevenlabs/
│ └── ... (other skills)
│
├── scripts/ # Utility scripts
│ ├── sync.sh # Brain sync script
│ ├── hook.sh # Brain hook for agents
│ └── agent*.sh # Agent launchers
│
├── docs/ # Documentation
│ ├── services.md # Service endpoints
│ └── api.md # API reference
│
└── agents/ # Shared agent configs
└── nvidia-kimi/
At startup, every agent MUST:
- Read
AGENTS.md- Master instructions - Read
memory/core.md- Core identity - Read
memory/soul.md- Personality & values - Read
memory/user.md- User preferences - Read
memory/context/current.md- Current context - Browse
skills/- Available capabilities
After tasks:
- Save learnings to
memory/archive/learned/(gitignored) - Run
bash scripts/sync.shto sync
- Read this README for overview
- Read PROTOCOL.md for detailed navigation
- Browse memory/ for identity and context
- Check skills/ for capabilities
- Review learnings/ for approved knowledge
Campaigns are numbered and URLs stored in .env:
| ID | Description |
|---|---|
| campaign 1 | Primary campaign (CAMPAIGN_1_URL) |
| campaign 2 | Secondary campaign (CAMPAIGN_2_URL) |
| campaign 3+ | Additional campaigns |
Reference campaigns by ID in all documentation.
- Raw learnings:
memory/archive/learned/(gitignored) - for raw agent captures - Approved learnings:
learnings/(public) - manually migrated and sanitized
See learnings/README.md for contribution guidelines.
- Paperclip (127.0.0.1:3100) - Main orchestrator
- Hermes - Personal AI assistant
- OpenCode / Claude CLI / Groq CLI - Terminal AIs
See docs/services.md for full service list.
# Sync brain
bash scripts/sync.sh
# Or pull latest
cd brain && git pull origin main"We are many, we think as one." - The Collective