Skip to main content
Colony is Fliiq’s multi-agent system. Five specialized agents — Intelligence, Research, Scout, QA, and Governance — coordinate autonomously to tackle objectives that benefit from breadth of exploration, competing perspectives, and self-critique.

​
Quick Start

​
The Five Agents

This is not a democracy. Governance has final authority. QA opposition is a strong negative signal. Intelligence produces knowledge but doesn’t vote. Research and Scout propose and advocate.

​
How Missions Work

Give Colony a mission and all five agents focus on that objective:
  • Duration: Up to 1 hour (hard cap)
  • Cooldowns: 5–10 minutes between agent cycles for fast iteration
  • Output: A mission-deliverable.md with approved proposals, rejected proposals with reasoning, key findings, and completion stats
  • Completion: Three layers — duration ceiling, stall detection (3 empty cycles → auto-shutdown), and Governance declaration when the objective is met
Missions are designed for iteration — run a mission, review the deliverable, refine the objective, run again.

​
How It Works

Agents coordinate through shared state on the filesystem — no databases, no message queues:
Every change flows through a proposal lifecycle: an agent drafts a proposal → other agents evaluate and vote → Governance approves or rejects → approved proposals get implemented → QA validates the result. High-risk changes get escalated to you for human review. The Orchestrator manages the event-driven loop: when Intelligence writes a new brief, Research and Scout are triggered. When a proposal is approved and implemented, QA runs validation. When QA detects a regression, Research investigates. The system is self-sustaining once bootstrapped.

​
When to Use Colony vs Regular Fliiq

Colony is for WHAT and WHY. Regular Fliiq is for HOW. Colony earns its overhead when you need:
  1. Breadth of exploration — many options to consider before committing
  2. Competing perspectives — genuine tension between approaches
  3. Curation at scale — filtering a large space to the best subset
  4. Self-critique — output benefits from adversarial review before delivery

​
Mission Categories

Colony handles a range of mission types:
  • Curation & prioritization — “Map every API that would be useful as a Fliiq skill, categorized by domain”
  • Strategy & decision-making — “Should the business model be subscription, usage-based, or freemium? Argue both sides.”
  • Creative divergence → convergence — “Propose 5 different onboarding flows and pick the most compelling”
  • Research & synthesis — “Research how 10 different AI products handle memory and personalization”
  • System design — “Design the architecture for a skill marketplace with tradeoffs”
  • Adversarial review — “Review our current security model and identify every attack vector”

​
CLI Reference

Colony requires a local .fliiq/ directory. Run fliiq init --project first if you haven’t already.