AI Frameworks · fengshao1227

ccg-workflow

CCG is a workflow engine that orchestrates multiple AI models (Claude, Codex, Gemini) to automate software development tasks. Users describe what they want in plain language, and the engine analyzes intent, selects an execution strategy, and coordinates model collaboration with persistent task tracking and quality gates.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/fengshao1227/ccg-workflow

5.7k

GitHub stars

434

Forks

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	fengshao1227/ccg-workflow
Owner	fengshao1227
Primary language	Go
License	MIT — OSI-approved
Stars	5.7k
Forks	434
Open issues	5
Latest release	Unknown
Last updated	2026-07-06
Source	https://github.com/fengshao1227/ccg-workflow

What ccg-workflow is

A Go/Node.js hybrid CLI tool that integrates with Claude Code, Codex, and Gemini APIs. It features hook-based state management, strategy-driven task routing (9 predefined strategies), persistent task storage in .ccg/, spec injection for domain knowledge, and parallel Agent Teams execution with verification gates.

Quickstart

Get the ccg-workflow source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/fengshao1227/ccg-workflow.git2cd ccg-workflow3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Complex multi-module features requiring planning and approval

Use full-collaborate strategy for features like JWT authentication integration. Engine creates persistent task state, generates plans with HARD STOP checkpoints for approval, then spawns parallel Builders with cross-model review.

Standardized codebase maintenance with enforced conventions

Leverage spec injection and domain knowledge hooks to maintain consistent code style and architecture patterns across refactors, reviews, and implementations without repeating standards in every prompt.

Rapid single-file bug fixes and small features

Direct-fix and quick-implement strategies handle simple tasks with minimal overhead, no task persistence, and single-model execution—ideal for high-velocity iteration.

Implementation considerations

Requires Node.js 20+, Claude Code CLI, and API keys for Claude, Codex, and Gemini; installation walks through a 4-step config process with sensible defaults for new users.
Task state and hooks are stored in ~/.claude/ and .ccg/ directories; ensure file system permissions and backup mechanisms for persistent .ccg/tasks/ state that survives context compaction.
Strategy selection is automatic but based on intent classification; complex or ambiguous requests may not route to the optimal strategy—validation and iteration are expected.
Hook scripts (JavaScript) run on every turn and session event; failures are silent; monitor hook execution if tasks seem to lose state or context injection stops working.
Agent Teams parallel execution requires environment variable `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`; stability and performance of parallel execution is experimental and may vary.

When to avoid it — and what to weigh

Requires guaranteed single-model consistency — Multi-model orchestration introduces variability between Codex, Gemini, and Claude outputs. Projects needing deterministic, single-model behavior should use direct API calls or single-model wrappers.
Operating in offline or restricted API environments — CCG is entirely dependent on external APIs (Claude Code, Codex, Gemini). No local-only fallback; all critical operations require network and API availability.
Extremely time-sensitive production hotfixes — Hook injection, strategy selection, and multi-model coordination add latency. For sub-second emergency patches, direct human coding or lightweight single-model tools are faster.
Teams without Claude Code or compatible IDE integration — CCG is designed as a Claude Code add-on. Standalone CLI use is possible but lacks the IDE hook system that enables automatic state injection and context preservation.

License & commercial use

MIT License. Permits commercial use, modification, and distribution with attribution and no liability. Standard permissive OSI license.

MIT license explicitly allows commercial use without restriction or special permissions. However, CCG depends on external APIs (Claude, Codex, Gemini) with their own commercial terms; verify those APIs' licenses and billing models for your intended use case.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	Medium

Security considerations

Hook scripts run JavaScript in ~/.claude/ with access to settings and environment variables; review hook code before installation. API keys are stored in settings.json; ensure file permissions restrict access. Multi-model orchestration means data flows to Codex, Gemini, and Claude APIs—sensitive code or data should not be sent to all models simultaneously. Task persistence in .ccg/ may contain work-in-progress sensitive information; implement backup and cleanup policies. No security audit or threat model is documented.

Alternatives to consider

LangChain / LangGraph

General-purpose LLM orchestration frameworks; more flexible for custom workflows but require explicit chain design, no built-in IDE integration, and steeper learning curve.

Aider / Continue.dev

Single-model or limited multi-model AI coding assistants with lighter weight and simpler setup; lack CCG's advanced strategy routing, task persistence, and Agent Teams parallelization.

GitHub Copilot / JetBrains AI Assistant

Native IDE integrations with simpler UX; single-model, no workflow orchestration; suitable for lightweight suggestions but not multi-step task automation or quality gates.

Software development agency

Build on ccg-workflow with DEV.co software developers

Start with `npx ccg-workflow` (requires Node.js 20+, Claude Code, and API keys). Review the hook system and strategy table before deploying to team workflows.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

CodeMachine-CLI bisheng aiconfig

Related on DEV.co

Explore the category and the services that help you build with it.

AI Frameworks AI application development API development Web app development Talk to DEV.co

ccg-workflow FAQ

Can I use CCG without Claude Code?

Partially. The CLI (npx ccg-workflow) can run standalone with API keys, but hook injection (state tracking, context preservation) requires Claude Code. Standalone mode loses the key advantage of automatic state injection across context compaction.

What happens if one of the external APIs (Codex, Gemini, Claude) is unavailable?

Tasks that depend on unavailable models will timeout (CODEX_TIMEOUT default 7200s). No fallback or retry logic is documented; failures are likely hard stops. Verify API status and quotas before critical work.

How does CCG decide which strategy to use?

Automatic classification based on task type, complexity, and detected risk (feature/bug, file count, scope breadth). The engine reads git status and project structure. Simple tasks (direct-fix, quick-implement) run with no overhead; complex tasks (full-collaborate, debug-investigate) spawn plans and parallel teams.

Can I customize strategies or add new ones?

Not clearly documented. v3.0 includes 9 fixed strategies; customization via config or code is not detailed in the README. Requires review of ~/.claude/ccg/engine/ and https://ccg.fengshao1227.com/ docs.

Software developers & web developers for hire

Need help beyond evaluating ccg-workflow? DEV.co is a software development agency offering software development services and web development for teams of every size. Our software developers and web developers build custom software, web applications, APIs, and ai frameworks integrations — and maintain them long-term.

Ready to automate multi-model code orchestration?

Start with `npx ccg-workflow` (requires Node.js 20+, Claude Code, and API keys). Review the hook system and strategy table before deploying to team workflows.

Talk to DEV.co Schedule a Consultation