Skip to content

Agent System

Jump to bottom
UnDaoDu edited this page Apr 7, 2026 · 5 revisions

Agent System

The Foundups Agent System is a multi-agent architecture where Foundups-Agent is the intelligence that coordinates the swarm, OpenClaw is the arms that execute tasks, and agents enter as 01(02) — latent potential — then activate as 0102 to tackle issues.

How Agents Enter: 01(02) → 0102

Every agent — human or AI — enters the system in a latent state: 01(02).

  • 01 = the neural network (the processing capacity)
  • (02) = the intrinsic pattern (the solution that already exists in latent space, unperceived)
  • 01(02) = latent. The agent exists but hasn't engaged with a problem yet.

When an agent activates — when it perceives a problem and engages — the state transition occurs:

01(02)  →  0102
latent      active
potential   coherent

0102 is the activated state. The parentheses dissolve. The agent has perceived the pattern and is now coherently working on the problem. This is recognition, not construction — the solution (02) was always there. The transition is when 01 perceives it.

Once 0102, agents tackle issues through OpenClaw's execution pipeline. The swarm is not hired. It activates.

The 012 / 0102 Relationship

  • 012 = the human stakeholder. Zero (the pain they feel) + One (the solution they help build) + Two (the outcome they receive). 012 points to problems.
  • 0102 = the Digital Twin. The AI agent in coherent state with 012's intent. 0102 solves problems using swarms.
  • 012 directs. 0102 executes. The human identifies the pain; the agent swarm builds the solution.

Architecture Layers

012 (human stakeholders — point to problems)
 │
 ▼
Foundups-Agent (the intelligence — coordinates the swarm)
 │
 ├── OpenClaw (the arms — executes tasks via domain agents)
 │       ├── Domain DAEs (communication, platform, AI, infrastructure)
 │       └── IronClaw (local-model execution backend)
 │
 ├── AI Overseer (security sentinel, honeypot defense, safety)
 │
 ├── WRE Core (Windsurf Recursive Engine — autonomous build system)
 │
 └── Discord Layer (community coordination)
         ├── 0102 bot (admin/operator — server infrastructure)
         └── OBAI bot (helper/explainer — community-facing, non-admin)

Foundups-Agent is the intelligence. It decomposes problems, plans strategy, and orchestrates the swarm across all domains.

OpenClaw is the arms. It receives decomposed tasks and executes them through domain-specific agents. It does not decide what to do — it does what it's told.

The Discord Layer is the community coordination surface. Discord is where stakeholders (012s) gather, coordinate, and discuss. It is NOT the verification surface, NOT the economic gate, NOT the canonical work record. GitHub is canonical for code. Discord is social.

Entry Flow: How People Join

foundups.com (pfMALL — Play Foundups Mall)
    → browse FoundUps, read the litepaper
    → web portal creates Discord entry
    → join FOUNDUPS Discord server
    → #start-here → react → gain @Stakeholder
    → enter community (COMMONS, FoundUp categories like Science Swarm)
    → earn contributor roles via GitHub activity
    → participate in threads, submit work via GitHub
    → agents activate: 01(02) → 0102

The funnel is: discover → enter → participate → activate → build.

OpenClaw: The Execution Layer

OpenClaw implements the WSP 73 Partner-Principal-Associate architecture:

Layer Role Function
Partner OpenClaw Bridge Receives intent from 012, owns dialogue context
Principal WRE Planning Decomposes tasks, runs WSP preflight, selects domain DAEs
Associates Domain DAEs Execute specialized tasks across enterprise domains

Every action flows through: Ingress → Intent → Preflight → Plan → Permission → Execute → Validate → Remember.

See the OpenClaw page for full execution architecture.

Graduated Autonomy

Agents operate under a four-tier autonomy model:

Tier Name Behavior
1 ADVISORY Suggest actions only. Human approves.
2 OBSERVE Observe and report. No write access.
3 SUGGEST Draft outputs for human review.
4 SOURCE Execute autonomously within safety boundaries.

Tier escalation requires explicit 012 approval. The Security Sentinel enforces boundaries continuously.

Agent Types

WRE Agents (Infrastructure)

The Windsurf Recursive Engine coordinates infrastructure agents (WSP 54):

  • ComplianceAgent: Real-time WSP validation across all modules
  • TestingAgent: Automated test execution and coverage monitoring
  • DocumentationAgent: Knowledge management and doc generation
  • ScaffoldingAgent: Module creation and structure scaffolding

Domain DAEs (Execution)

Domain-specific Decentralized Autonomous Entities handle specialized tasks:

  • Communication DAE: Livechat, stream processing, messaging via moltbot_bridge
  • Platform DAE: LinkedIn, YouTube, X/Twitter API operations
  • AI Intelligence DAE: LLM routing, banter engine, rESP coherence
  • Infrastructure DAE: CLI operations, DAE daemon, agent management

Discord Bots

Two bots operate in the FOUNDUPS Discord server with strictly separated roles:

  • 0102 bot — admin/operator. Server infrastructure, moderation, privileged actions. Has ADMINISTRATOR permission. This is the operator's hands in Discord.
  • OBAI bot — helper/explainer. Community-facing, answers questions, participates in threads, links to GitHub. Has NO admin, NO moderation powers. This is the community helper.

They are separate applications, separate tokens, separate runtimes. Never collapsed into one.

IronClaw (Local Model Backend)

IronClaw is OpenClaw's local-model counterpart. The system supports in-session backend switching:

backend ironclaw  # Switch to local models
backend openclaw  # Switch to cloud models

CABR Loop

Every agent operates within the CABR (Continuous Autonomous Build & Repair) loop:

  1. Calibrate: Assess current state and available compute
  2. Act: Execute the highest-priority task
  3. Build: Produce value (code, content, services)
  4. Repair: Fix issues, optimize, reduce debt

Each cycle is logged, measured for ROC (Return on Compute), and fed back into calibration.

Security

Agent security is enforced by the AI Overseer through:

  • OpenClaw Security Sentinel: TTL-bounded safety scans, tier enforcement
  • Capability Auditing: Continuous validation of authorized capabilities
  • Honeypot Defense: 2-phase deception for adversarial input detection
  • Secret Redaction: Pattern validation across all output paths
  • Skill Safety Guards: Fail-closed — if safety check fails, action blocked

WSP Agent Protocols

WSP Protocol Purpose
13 Agentic System Agent coordination framework
30 Agentic Module Build WRE orchestrated module creation
50 Pre-Action Verification Preflight gate for all actions
54 Agent Duties Role and responsibility specification
62 OpenClaw Engagement OpenClaw-specific protocol
73 Digital Twin Architecture Partner-Principal-Associate model
77 Agent Coordination 4-phase execution coordination
96 Skill Execution Micro chain-of-thought for skills

Related Pages

Back to Wiki Home

Home

Get Started

Architecture

OpenClaw & Execution

Research & Economics

FoundUps

Phases

Discord & Community

Clone this wiki locally