DannyMac180/fable-advisor

Claude Fable as an orchestrator for Opus, GPT and Grok

What it solves

It implements an "architect pattern" for AI-assisted coding to optimize the balance between reasoning quality, token cost, and accuracy. It prevents the "blind spots" that occur when a single model family handles both implementation and review by routing tasks across different model vendors (Claude and OpenAI) and varying reasoning effort levels based on the complexity of the task.

How it works

The project is a plugin for Claude Code that establishes a multi-agent orchestration system:

  • The Architect (Session Model): Runs on Fable 5.1. It handles high-level requirements, decomposes tasks into specifications, and verifies the final output.
  • Implementation Lanes: Tasks are routed to specific agents based on complexity:
    • Routine: Routed to codex-implementer (GPT-5.6 Luna) for straightforward tasks where the spec fully determines the outcome.
    • High-complexity: Routed to sol-implementer (GPT-5.6 Sol) for tasks requiring nuanced judgment, such as security-sensitive paths or complex refactors.
  • The Advisor: A read-only agent (Fable 5.1) that operates in a clean context to provide a skeptical final review of the accumulated changes before they are shipped.
  • Dynamic Effort: The architect specifies the reasoning effort (low to ultra) for each task, ensuring expensive compute is only used where it actually changes the outcome.

Who it’s for

Developers using Claude Code who want a rigorous, multi-model workflow for high-stakes software engineering where architectural integrity and cross-vendor verification are critical.

Highlights

  • Cross-Vendor Review: Automatically integrates Claude and GPT models to eliminate single-model family blind spots.
  • Context-Clean Skepticism: Uses a dedicated advisor agent in a fresh context to prevent the session's accumulated assumptions from biasing the final review.
  • Reasoning Effort Routing: Allows the architect to specify reasoning levels per task rather than pinning a global effort level.
  • Orchestration Skill: Provides a routing doctrine and a six-part spec contract to make context-free delegation safe.

Related

  • Project
    hex/claude-council

    A Claude Code plugin that queries multiple AI coding agents in parallel to provide side-by-side comparisons and a synthesis of consensus and divergence for technical decisions.

  • Project
    Nanako0129/pilotfish

    pilotfish is a policy‑driven orchestration layer for Claude Code that automatically delegates low‑effort coding tasks to cheaper Claude models (Haiku, Sonnet) while keeping high‑level planning and final judgment in a main Opus session. It installs via a macOS/Linux plugin or a legacy global config, defines role agents in markdown, and uses a dispatch flow to route work based on interaction shape and risk. The project includes documentation, benchmarks, and a MIT license.

  • Project
    Hainrixz/the-architect

    A Claude Code meta-agent that designs rigorous software blueprints with verified stacks and strict acceptance criteria to guide autonomous AI builders.

  • Project
    Cjbuilds/Codex-Orchestration

    Codex Orchestration is a Codex plug‑in that lets you attach other LLMs (Claude Fable 5, Opus 5, OpenRouter models, etc.) to a task and assign them specific roles—Planner, Advisor, Designer, Executor. Codex stays the top‑level orchestrator, iterating plans, reviewing them, optionally designing UX artefacts, and finally executing code. The plug‑in is installed via the Codex marketplace, configured with skill prompts inside a Codex chat, and supports role‑specific effort levels. It aims to improve plan quality, enable parallel work, and reduce premium‑model usage. MIT‑licensed.

  • Project
    Yeachan-Heo/oh-my-claudecode

    A multi-agent orchestration framework for Claude Code that enables natural-language task distribution across specialized agents and multiple AI providers.