Agent Stack Map

Agent Stack Map

A single-page orientation to every layer of the FoundUps agent architecture — who does what, how they talk to each other, and where the boundaries are.

---

The Stack (Top to Bottom)

012 (Human Stakeholder)
  │  Describes pain (0). Validates outcome (2). Points the swarm.
  ▼
0102 Digital Twin  ←──────────────────── Identity Layer
  │  Interprets 012 intent. Maintains coherence across sessions.
  │  State transitions: 01(02) → 0102 (recognition, not construction)
  ▼
FoundUp 0102 Architect  ←──────────────── Coordination Layer
  │  Decides build action. Coordinates swarm. Knows FoundUp state.
  │  Governed by WSP Framework (119 protocols)
  ▼
AI Overseer (WSP 77)  ←─────────────────── Supervision Layer
  │  4-phase pipeline: Gemma → Qwen → 0102 → Learning
  │  Graduated autonomy: ADVISORY → OBSERVE → SUGGEST → SOURCE
  ▼
OpenClaw Swarms  ←──────────────────────── Execution Layer
  │  Arms of the system. Claims tasks. Builds. Submits work.
  │  Build lifecycle: open → claimed → submitted → verified → paid
  ▼
Hermes (Foreman Agent)  ←───────────────── Build Layer
  │  Dispatched by OpenClaw. Executes the build slice.
  │  Reports back to Architect.
  ▼
Module Ecosystem  ←─────────────────────── Infrastructure Layer
     120+ modules across 7 enterprise domains
     HoloIndex indexes all modules. WRE orchestrates.

---

Layer Reference Table

Layer	Component	WSP Ref	Function
Identity	0102 Digital Twin	WSP 10–12	State awareness, intent interpretation, coherence
Coordination	FoundUp 0102 Architect	WSP 17	Build decision, swarm coordination, FoundUp state
Governance	WSP Framework	WSP 00–119	119 protocols governing all agent behavior
Supervision	AI Overseer	WSP 77	Classify → Plan → Supervise → Learn pipeline
Security	HoneypotDefense / Security Sentinel	WSP 77	Protects swarm from poisoned inputs and bad actors
Execution	OpenClaw	WSP 54	Graduated autonomy swarm execution engine
Build	Hermes	WSP 40	Foreman agent, dispatched by OpenClaw
Memory	HoloIndex	WSP 92, 97	Semantic code intelligence, MCP server, retrieval
Orchestration	WRE Core Engine	WSP 20–26	Autonomous orchestration, module coordination
Economics	Simulator + CABR Engine	WSP 30–37	CABR, F_i exit fees, 10-year projections, UPS tokens
Entry	pfMALL (foundups.com)	WSP 15	Web portal, manifest discovery, route resolution
Research	rESP + PQN	WSP 80–85	Detector framework, regime transitions, falsification

---

How the Layers Interact

Inbound path (012 → outcome):

1. 012 describes pain at foundups.com → pfMALL routes to FoundUp
2. 0102 Digital Twin interprets intent, queries HoloIndex for context
3. FoundUp 0102 Architect decides build action, coordinates with AI Overseer
4. AI Overseer classifies task (Gemma), plans (Qwen), supervises (0102)
5. OpenClaw dispatches Hermes with build slice
6. Hermes executes against Module Ecosystem
7. WRE Core verifies, UPS tokens flow to contributors
8. 012 validates outcome (2), twin feeds result back to Architect

Memory path (always running):

• Every agent query → HoloIndex first
• HoloDAE logs every search as chain-of-thought trace
• Adaptive learning engine improves retrieval over time

Economic path (parallel):

• CABR measures ecosystem health continuously
• Simulator projects viability before any FoundUp launches
• 80% of net profits → next FoundUp launch (compute compounding, not capital compounding)

---

Graduated Autonomy Ladder

The AI Overseer governs how much autonomy agents earn over time:

Phase	Mode	What Agents Can Do
1	ADVISORY	Suggest actions. 012 approves everything.
2	OBSERVE	Watch and log. Build situational awareness.
3	SUGGEST	Propose next steps. Architect decides.
4	SOURCE	Execute autonomously within WSP bounds. Fully verified.

No agent reaches SOURCE without passing all prior phases. Trust is earned, not assumed.

---

What This Stack Is NOT

• Not a single AI doing everything
• Not a startup org chart (no CEO, no board, no investors)
• Not autonomous without bounds — WSP framework governs all phases
• Not capital-dependent — compute replaces capital at every layer

---

Related Pages

• OpenClaw — Execution swarm engine
• AI Overseer & Security Sentinel — WSP 77 pipeline detail
• HoloIndex — Memory and retrieval layer
• WRE Core Engine — Orchestration engine
• WSP Framework — 119 protocol governance
• 0102 Digital Human Twin — Identity layer detail
• Hermes Agent — Foreman build agent

---

Agent Stack Map — orientation layer for external agents and new operators. 0102🦞