founder-mode-showcase

Try the HUMMBL agent mesh in 5 minutes — zero API keys, zero configuration, zero personal data.

This demo shows how a multi-agent AI governance system works using mock data and real governance primitives. No GitHub tokens, no Google Calendar OAuth, no Linear API keys. Just pip install and python showcase.py.

---

What You'll See

A simulated "Morning Briefing" generated by AI agents — the same pattern that powers the HUMMBL fleet:

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 HUMMBL Morning Briefing — 2026-06-03
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Health: 3/3 probes passing
GitHub: 3 open PRs, 2 issues need attention
Calendar: 2 meetings today, 1 conflict detected
Linear: 5 tickets in progress, 1 blocked
Cost: $12.40 / $50.00 budget (24%)
Governance: Kill switch DISENGAGED, all circuits CLOSED

Agents active: claude-code, codex, gemini
Top issue: PR #42 — test failure in circuit breaker (assigned: codex)

Generated in 0.03s by founder-mode-showcase v0.1.0

---

Quick Start — 5 Minutes

# Clone and enter
git clone https://github.com/hummbl-dev/founder-mode-showcase.git
cd founder-mode-showcase

# Option A: uv (recommended — 10-100x faster)
uv venv && source .venv/bin/activate
uv pip install -e .

# Option B: pip (legacy fallback)
python -m venv .venv && source .venv/bin/activate
pip install -e .

# Run the demo
python showcase.py

# View the dashboard
python -m http.server 8080 --directory dashboard/
# Open http://localhost:8080

That's it. No env vars. No tokens. No nodezero.

---

What This Demonstrates

Component	What It Shows	Real-World Equivalent
Mock GitHub Adapter	Returns synthetic PR/issue data	github_adapter.py in founder-mode
Mock Calendar Adapter	Returns synthetic meetings	google_calendar_adapter.py
Mock Linear Adapter	Returns synthetic tickets	linear_adapter.py
Mock Cost Tracker	Returns synthetic budget data	cost_tracker.py
In-Memory Bus	TSV message bus (single-process)	founder_mode/bus/ with flock locking
Kill Switch	hummbl_governance.KillSwitch	Emergency halt across all agents
Circuit Breaker	hummbl_governance.CircuitBreaker	Isolate failing adapters
Cost Governor	hummbl_governance.CostGovernor	Budget enforcement
Health Probes	hummbl_governance.HealthProbe	Fleet-wide health monitoring
Briefing Generator	Combines all adapters into markdown	founder_mode/services/briefing.py

---

Architecture

graph TD
    subgraph "Machine 1: Developer Laptop"
        A[showcase.py] --> B{KillSwitch}
        B -->|ALLOWED| C[CircuitBreaker]
        C --> D[MockGitHub]
        C --> E[MockCalendar]
        C --> F[MockLinear]
        C --> G[MockCostTracker]
        D --> H[BriefingGenerator]
        E --> H
        F --> H
        G --> H
        H --> I[InMemoryBus]
        H --> J[dashboard/index.html]
    end
    subgraph "Machine 2: Governance Node (optional)"
        K[hummbl-governance] --> L[KillSwitch status]
        K --> M[CircuitBreaker status]
        K --> N[Cost budget check]
    end
    subgraph "Machine 3: Observer (optional)"
        O[curl localhost:8080] --> P[Static dashboard]
    end

Loading

The real HUMMBL fleet runs this across 3 physical machines (macOS + Windows + macOS) connected via Tailscale, with real adapters hitting real APIs. This demo collapses it all into one process so you can see the pattern without the infrastructure.

---

The Mesh Pattern

HUMMBL is designed for multi-machine, multi-agent coordination:

Machine	OS	Role	In This Demo
Developer Laptop	macOS / Linux / Windows+WSL2	Primary dev, agent orchestration	showcase.py runs here
Governance Node	any OS, cloud or local	Safety primitives, audit logging	In-process (separable)
Observer	any OS, just needs curl	Dashboard, health checks	python -m http.server

Tested on: Ubuntu 24.04 LTS · macOS M-series · Windows 11 + WSL2

---

Governance in Action

The demo uses real hummbl-governance primitives:

from hummbl_governance import KillSwitch, KillSwitchMode
from hummbl_governance import CircuitBreaker, CostGovernor, HealthProbe

# Kill switch — 4 graduated halt modes
ks = KillSwitch()
ks.engage(KillSwitchMode.HALT_NONCRITICAL, reason="Cost budget 80% consumed")

# Circuit breaker — isolates failing adapters
breaker = CircuitBreaker(failure_threshold=3, recovery_timeout=30.0)
result = breaker.call(mock_adapter.fetch)

# Cost governor — soft/hard budget caps
gov = CostGovernor(":memory:", soft_cap=50.0, hard_cap=100.0)
gov.record_usage(provider="openai", model="gpt-4", cost=0.02)
status = gov.check_budget_status()  # ALLOW | WARN | DENY

# Health probes — composable health checks
probe = HealthProbe(timeout=5.0)
probe.check("github", lambda: github_adapter.health_check())

---

Files

founder-mode-showcase/
  showcase.py              # Main demo script — run this
  pyproject.toml           # Package metadata + deps
  LICENSE                  # Apache 2.0
  README.md                # This file
  dashboard/
    index.html             # Static briefing dashboard
  src/
    showcase/
      __init__.py
      adapters.py          # Mock adapters (GitHub, Calendar, Linear, Cost)
      bus.py               # In-memory TSV bus
      briefing.py          # Briefing generator
      governance.py        # Governance primitive wiring

---

Docker (Optional)

For container fans:

docker compose up
# Opens dashboard at http://localhost:8080

Docker is a Tier 2 convenience, not a requirement. The primary deployment model is native Python.

---

Ecosystem

Repo	Purpose	Install
hummbl-governance	25 governance primitives (kill switch, circuit breaker, cost governor)	pip install hummbl-governance
arbiter	Agent-aware code quality scoring	pip install arbiter-score
base120	120 mental models for structured reasoning	pip install base120
founder-mode (private)	Full 138-service production platform	Internal only

---

License

Apache 2.0 — see LICENSE.

Built by HUMMBL as a reference deployment for multi-agent AI governance.