ccg-workflow: a multi-model orchestrator for Claude Code that delegates specialized work to Codex and Gemini

What it solves

CCG is a workflow engine designed to enhance Claude Code, transforming it from a single-model agent into a multi-model orchestrator. It solves the problem of relying on a single LLM for complex software engineering tasks by allowing Claude to delegate specialized work to other models like Codex (OpenAI) and Gemini (Google) for parallel analysis, implementation, and cross-review.

How it works

CCG uses a Go binary bridge (codeagent-wrapper) to connect Claude Code to external models. It employs a "Hook Engine" that injects state and project context into every turn of the session, ensuring that context is not lost during compaction. The system classifies tasks by complexity and automatically selects one of 10 built-in strategies (e.g., full-collaborate for complex features or direct-fix for simple bugs).

For complex tasks, it creates a persistent task directory (.ccg/tasks/) containing requirements, plans, and review results. It also includes "Quality Gates" to verify security, quality, and documentation sync before completion.

Who it’s for

Developers using the Claude Code CLI who want to automate complex multi-step software development workflows and leverage the strengths of multiple AI models simultaneously.

Highlights

• Multi-Model Orchestration: Dispatches work to Codex and Gemini in parallel for higher quality results.
• Hook Engine: Prevents context loss through automatic state and domain knowledge injection.
• Strategy-Based Execution: 10 predefined strategies that scale from simple fixes to full-scale collaboration.
• Persistent Task System: Maintains a detailed lifecycle for medium-to-high complexity tasks.
• Domain Knowledge Base: Over 100 domain-specific files (e.g., Security, DevOps, AI/MLOps) that are auto-injected based on keywords.
• Quality Gates: Built-in verification steps for security and code quality.