Fable Council

Convene a council of AI coding agents before finalizing an architectural plan. Propose → hear independent assessments from each installed agent CLI → revise, repeated across bounded deliberation rounds.

Purpose

Enforce multi-agent deliberation and cross-model consensus on complex plans. Leverage complementary models (Claude, Codex, Cursor, Gemini, Grok, OpenCode) to detect blind spots, challenge flawed assumptions, and converge on robust engineering plans.

When to Use

• When planning high-stakes architectural changes, major refactorings, or migrations.
• When the user asks to convene the council or get second opinions ("ask the council", "consult the other agents").
• When deliberating between multiple competing implementation strategies.

When NOT to Use

• Bounded single-file bugfixes or routine implementation tasks (use fable-execute or fable-tdd).
• When no other agent CLIs are installed on the local system (fall back to solo fable-plan).
• Post-implementation diff verification or code review (use fable-review or fable-judge).