Route tasks to 9 AI providers, orchestrate parallel workflows, synthesize results
14 skills, 4 orchestration patterns, 3 executable templates
Modules • Templates • Patterns • Quick Start
English | 中文
Multi-AI Coordination Toolkit enables Claude Code to delegate tasks to 9 AI providers running in separate terminal sessions. Beyond simple delegation, it provides orchestration patterns and templates for parallel fan-out, consensus voting, sequential pipelines, and specialist routing.
| Before (v1) | After (v2) |
|---|---|
| Flat list of 14 skills | Organized by function: delegation, monitoring, workflows |
| No orchestration guidance | 4 orchestration patterns with step-by-step instructions |
| No templates | 3 executable templates for common multi-AI workflows |
| Skills-only | Router SKILL.md auto-routes based on user intent |
| Provider | Command | Best For |
|---|---|---|
| Codex | /cask <msg> |
Code generation, refactoring, PR review |
| Gemini | /gask <msg> |
Long-context analysis, research, multimodal |
| OpenCode | /oask <msg> |
Code analysis, debugging, architecture |
| DeepSeek | /dskask <msg> |
Reasoning, math, algorithms |
| iFlow | /iask <msg> |
Workflow automation, integrations |
| Kimi | /kask <msg> |
Long documents, Chinese-language tasks |
| Qwen | /qask <msg> |
Multilingual, general-purpose analysis |
| Droid | /dask <msg> |
Mobile/Android development |
| Generic | /ask <provider> <msg> |
Any provider by name |
| Skill | Command | Description |
|---|---|---|
| pend | /pend <provider> |
Fetch latest reply from a provider |
| ping | /ping <provider> |
Check provider connectivity |
| Skill | Command | Description |
|---|---|---|
| all-plan | /all-plan |
Multi-AI collaborative planning |
| stem-modeling | /stem-modeling |
STEM academic modeling |
| ccb-launcher | /ccb-launcher |
CCB environment management |
Ready-to-use templates for common multi-AI workflows:
| Template | Use Case |
|---|---|
| parallel-review | Send code review to multiple providers in parallel, then synthesize |
| consensus-decision | Get multiple AI opinions on a design/architecture decision |
| research-delegation | Delegate research tasks to specialized providers |
For advanced multi-AI coordination, see orchestration/patterns.md:
| Pattern | When to Use |
|---|---|
| Parallel Fan-Out | Same task to N providers, synthesize results |
| Sequential Pipeline | Chain providers: output of one feeds the next |
| Consensus Voting | Multiple providers vote on a decision |
| Specialist Routing | Route sub-tasks to the best-fit provider |
cd ~/.claude/skills
git clone https://github.com/LeoLin990405/claude-ccb-skills.git# The toolkit auto-routes based on your request:
"Review this code with Gemini and Codex" --> parallel fan-out + parallel-review template
"Get opinions on this architecture" --> consensus-decision template
"Ask DeepSeek to solve this algorithm" --> /dskask
# Direct provider delegation:
/gask "analyze this codebase"
/cask "review this PR"
# Check results:
/pend gemini
/ping codex多 AI 协调工具集 让 Claude Code 能够将任务委托给 9 个独立终端会话中运行的 AI 服务商。除了简单委托外，还提供编排模式和模板，支持并行扇出、共识投票、顺序流水线和专家路由。
| 类别 | 技能数 | 覆盖范围 |
|---|---|---|
| 委托 | 9 | Codex、Gemini、OpenCode、DeepSeek、iFlow、Kimi、Qwen、Droid、通用 |
| 监控 | 2 | 获取回复 (pend)、连接检查 (ping) |
| 工作流 | 3 | 多 AI 规划、STEM 建模、环境管理 |
| 模板 | 用途 |
|---|---|
| parallel-review | 并行代码审查，多服务商同时审查后综合结果 |
| consensus-decision | 多 AI 共识决策，获取多方意见后投票 |
| research-delegation | 研究任务委托，按专长分配给最合适的服务商 |
| 模式 | 适用场景 |
|---|---|
| 并行扇出 | 同一任务发给多个服务商，综合结果 |
| 顺序流水线 | 链式调用，前一个的输出作为下一个的输入 |
| 共识投票 | 多个服务商投票决定 |
| 专家路由 | 按子任务特点分配给最擅长的服务商 |
cd ~/.claude/skills
git clone https://github.com/LeoLin990405/claude-ccb-skills.git- Leo (@LeoLin990405) - Project Lead
- Claude (Anthropic Claude) - Content Generation
MIT License - see LICENSE for details.
Built with collaboration between human and AI