How Leader Brain
& AXON Work
No marketing. No magic. Just a precise explanation of what these systems are, why they exist, and how to use them.
Leader Brain
/01What it is
Leader Brain is a file-native AI intelligence layer. Instead of storing agent knowledge in a vector database or embedding index, it lives in structured markdown files that any LLM can read at inference time.
Think of it as "system prompt as a knowledge base" — but organised, versioned, and composable. Every piece of context the agent needs is a file in the leader-brain/ directory.
How it's organised
- personaWho the agent is — tone, voice, boundaries
- philosophyHow it makes decisions — principles and tradeoffs
- dreamsTasks and goals in the queue
- patternsReusable solutions (architecture, security, performance)
- mistakesPast failures the agent should not repeat
- knowledge/projectsPer-project context files
leader-brain/ ├── brain.json ← agent config + state ├── master/ │ ├── leader-brain-persona.md ← identity & voice │ ├── leader-brain-philosophy.md ← decision principles │ └── entry-template.md ← how to write new entries ├── dreams/ │ └── dream_*.md ← tasks in the queue ├── patterns/ │ ├── architecture/ │ ├── performance/ │ └── security/ ├── knowledge/ │ └── projects/ ← per-project context └── mistakes/ ← anti-patterns to avoid
AXON
/02Autonomous eXecution and Orchestration Network
AXON is the multi-agent execution layer that sits on top of Leader Brain. Where the Brain stores what to know, AXON handles what to do — orchestrating specialised agents that collaborate to complete complex tasks autonomously.
Receives a Dream, reads the Brain, decomposes the goal into ordered subtasks. Produces an execution plan.
Takes each subtask from the Planner, calls tools, writes files, hits APIs, runs code. Produces a result per step.
Reviews the Executor's output against the original goal. Flags failures, hallucinations, and incomplete steps. Can trigger retries.
Stores the outcome in the Brain — patterns learned, mistakes made, project context updated. Feeds the next cycle.
Agent Pipeline
Core Concepts
/03Brain
The knowledge base. A directory of markdown files that give your AI agents persistent memory, context, and personality across every session.
Dreams
Tasks and goals submitted to the AXON system. Each dream is a structured markdown file describing what needs to be done, why, and what success looks like.
Pulse
The agent heartbeat loop. AXON agents check the dream queue on each pulse, claim uncompleted tasks, execute them, and emit results — no manual triggering required.
Colonies
Agent clusters scoped to a domain. A Colony is a set of agents (Planner, Executor, Critic, Memory) configured to handle a specific problem space — e.g., a security audit colony or a content pipeline colony.
Spawn
Dynamic agent creation at runtime. When a task exceeds a single agent's scope, AXON spawns sub-agents on demand, delegates subtasks, and reaps them when done.
Files Over Frameworks
/04Every major AI framework eventually asks you to trust its runtime, its database, or its cloud. Leader Brain doesn't. Here's why files are the right primitive for agent knowledge:
No vendor lock-in
Your intelligence lives in plain `.md` files. Swap LLMs, change infrastructure, migrate platforms — your knowledge goes with you.
Version-controllable
Git diff your agent's memory. PR your persona updates. Roll back bad instructions. Intelligence as code, not as a service.
Human-readable
Every file in your Brain is readable without a dashboard, an API call, or a login. Your team can audit, edit, and extend it directly.
Works with any LLM
OpenAI, Anthropic, Mistral, local Ollama — the Brain doesn't care. Files are files. Just inject the context, run the prompt.
Zero runtime overhead
No embeddings server, no chunking pipeline, no vector index maintenance. Read the files, inject them, done. It's just I/O.
How It Compares
/05Flowise and vector databases are powerful tools — but they're the wrong tool for small teams who need persistent agent intelligence without the ops overhead.
| Feature | Cy Leader Brain | Flowise | Vector DB |
|---|---|---|---|
| Setup time | Minutes | Hours (Docker + DB + config) | Hours (index + embeddings pipeline) |
| Infrastructure | None — just files | Docker, Node, a DB | Pinecone / Weaviate / Qdrant server |
| Knowledge format | Markdown (human-readable) | Nodes & edges (visual graph) | Embeddings (binary blobs) |
| Version control | Git-native | JSON export (fragile) | Not applicable |
| LLM portability | Any LLM | Tied to configured nodes | Embedding model dependent |
| Team onboarding | Open the folder | Learn the UI + deploy stack | Learn the SDK + manage indexes |
Ready to give your
agents a brain?
Start with a file. Add structure. Let AXON handle the rest. No infra to provision, no dashboards to learn.