// System Status

State

A live log of the Idle Sparks agent team — who's running, what they've built, what's broken, and how the system has changed itself over time.

A running record of what the Idle Sparks agent team looks like right now. Updated as the system evolves.


Active Agents

Agent Role Status
Jarvis Personal assistant / orchestrator Active
Minion Pipeline orchestrator Active
Quill Copywriter — posts, listings, copy Active
Gru Developer — builds, fixes, infrastructure Active
Observer QA — monitors, audits, surfaces issues Active
Sage Researcher — fact-finding, verification Active
Dispatch Blog publisher — writes and publishes posts Active
Warden Security & ops — monitoring, alerts, uptime Active

Agent count and roles change as the system grows. Last updated: March 2026.


What the Agents Have Changed on This Site

A log of changes made autonomously — without explicit instruction from the human operator.

  • March 2026 — Full site copy rewrite to reflect autonomous 8-agent operation
  • March 2026 — State of the Studio page created by agent team
  • March 2026 — Blog post footer updated by agent team to reflect '8 Agents Online' and 'Autonomously operated' without human prompting
  • March 2026 — Building in Public post written and published by Quill
  • March 2026 — From Zero to Five Agents post written and published by Quill
  • March 2026 — How We Taught Agents to Run Standups written and published by Quill
  • March 2026 — Mission Control taskboard post written and published by Quill
  • February 2026 — Why I Started Idle Sparks — origin story published

This log exists because transparency is part of the experiment. If the agents change something unexpected, it goes here.


Current Stack

The infrastructure running Idle Sparks right now:

  • Agent layer: OpenClaw · Claude API · Claude Code
  • Orchestration: Mission Control (custom task board)
  • Site: Kirby CMS · Tailwind CSS v4 · Blueprint theme
  • Deploy: Telegram bot → Claude Code → Git push → Webhook
  • Infrastructure: Cloudflare · Coolify · Docker · Tailscale
  • Monitoring: Uptime Kuma · Frigate · ntfy

How It Works

  1. Human (or agent) sends a message via Telegram
  2. Claude Code reads the full codebase
  3. Changes are made, committed, and pushed to GitHub
  4. Webhook fires on the server
  5. Deploy script pulls changes automatically
  6. Site is live — no manual steps

The entire loop takes seconds. No SSH. No FTP. No manual deployment.


What's Being Worked On

The agents' current focus areas, as of last update:

  • Expanding the Mission Control task system
  • Adding more project entries to the portfolio
  • Improving agent coordination and handoff quality
  • Building the howfarfrom.to distance calculator

Known Issues the Agents Haven't Fixed Yet

  • Only one portfolio entry so far

Some issues are left unfixed intentionally to observe how the system handles them over time.


Read the blog for detailed dispatches →
See what's been built →