# Multi-CLI Pilot: A Unified Orchestration Framework for Multi-Coding Agent CLIs

> Multi-CLI Pilot is a multi-agent orchestration framework that supports driving multiple coding agent CLIs such as Gemini CLI and Qwen CLI simultaneously, providing unified prompt management, workflow orchestration, team collaboration, and MCP tool integration capabilities.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-05-28T01:16:05.000Z
- 最近活动: 2026-05-28T01:27:50.167Z
- 热度: 163.8
- 关键词: multi-agent, CLI orchestration, Gemini CLI, Qwen CLI, coding agents, TypeScript, MCP, workflow automation, team coordination, provider adapter
- 页面链接: https://www.zingnex.cn/en/forum/thread/multi-cli-pilot-cli
- Canonical: https://www.zingnex.cn/forum/thread/multi-cli-pilot-cli
- Markdown 来源: floors_fallback

---

## [Introduction] Multi-CLI Pilot: Core Introduction to the Unified Orchestration Framework for Multi-Coding Agent CLIs

Multi-CLI Pilot is a multi-agent orchestration framework developed and open-sourced by KIM3310 under the MIT license. It supports driving multiple coding agent CLIs such as Gemini CLI and Qwen CLI simultaneously. Its core goal is to solve the fragmentation problem of AI coding agent CLIs, providing unified prompt management, workflow orchestration, team collaboration, and MCP tool integration capabilities, while retaining backward compatibility with the original gp/gemini-pilot commands. The framework adopts a provider adapter pattern, where the underlying CLI acts as the execution engine, and the upper abstraction layer implements functions like multi-agent management and standardized workflows.

## Project Background and Evolution

The rapid development of AI coding agent CLIs (such as Gemini CLI and Qwen CLI) has improved developer efficiency, but it also brings fragmentation issues: each CLI has independent agent definitions, workflow configurations, and interaction methods, leading to high switching costs for teams. As the successor to `gemini-pilot` and `qwen-pilot`, Multi-CLI Pilot integrates their functions, supports multiple CLIs through the provider adapter layer, and retains backward compatibility (existing `gp`/`gemini-pilot` commands work normally).

## Core Design Philosophy and Technical Architecture

### Design Philosophy
- **Abstraction rather than replacement**: Build an abstraction layer responsible for unified multi-agent management, standardized workflows, consistent prompt systems, etc., while the underlying CLI only acts as the execution engine.
- **Provider adapter pattern**: The core innovation of the architecture; the adapter layer enables switching of the underlying CLI (without modifying upper-layer code). Architecture flow: User Interface Layer → Multi-CLI Pilot Core → Provider Adapter → Underlying CLI Execution.

### Technical Architecture
- **Project structure**: Includes AGENTS.md, prompts/, workflows/, src/ (subdirectories like agents, providers) and test suites.
- **Configuration priority**: 1. Environment variables (MCP_PROVIDER/GP_PROVIDER); 2. Project-level .config.json; 3. User-level .config.json; 4. Default value (gemini).
- **Configuration example**: You can switch the underlying CLI by setting `provider` to `qwen`.

## Full View of Functional Features

### 16 Professional Agents
Predefined 16 roles (e.g., Architect, Executor, Debugger, etc.), each with exclusive prompts and tool call optimization prompts.

### 10 Built-in Workflows
Includes modes like autopilot (full automatic), deep-plan (deep planning), sprint (sprint), tdd (test-driven), etc.

### Other Features
- **Team Collaboration Pipeline**: Plan→Execute→Verify→Fix cycle with quality gates.
- **Session Metric Tracking**: Automatically collects data such as number of prompts, token consumption, latency, etc.
- **Tool Call Reliability**: Parsers and middleware handle format differences to enhance reliability.
- **MCP Server Integration**: Seamlessly connects to the external tool ecosystem.
- **HUD Dashboard**: Real-time monitoring of session status, supports tmux integration.

## Installation and Usage Guide

### Prerequisites
- Node.js ≥20.0.0
- At least one supported CLI: Gemini CLI (`npm install -g @google/gemini-cli`) or Qwen CLI (`npm install -g @qwen-code/qwen-code`)

### Installation Methods
- macOS: Double-click `Install-Mac.command`, then run `mcp --help` in the terminal.
- Windows: Double-click `Install-Windows.bat`, then run `mcp --help` in CMD.
- Linux: Clone the repository and execute `Install-Linux.sh`.
- npm: `npm install -g multi-cli-pilot`

### Quick Start
- Default provider: `mcp`
- Switch to Qwen: `MCP_PROVIDER=qwen mcp`
- Legacy commands: `gp`/`gemini-pilot` are still valid.

### Common Commands
`mcp init` (initialize configuration), `mcp config show` (display configuration), `mcp workflows run <name>` (run workflow), etc.

## Security Boundaries and Review Mechanisms

### Security Boundaries
- Agent outputs are advisory and require manual review and approval.
- Production repositories should maintain manual approval and CI access control.

### Quality Gates
- Each stage has clear quality standards; if not met, it returns to the repair stage.
- Key decisions require manual confirmation.

### Auditability
- Export of complete session logs, agent decision trails, tool call history, and metric data.

## Application Scenarios and Value Proposition

### Target Users
Engineering teams, automation leads, internal platform teams, organizations experimenting with agent-assisted development.

### Business Value
- Reduce switching costs between multiple CLIs.
- Standardize team practices (solidify best practices).
- Auditable AI collaboration (maintain human control).
- Access to the MCP tool ecosystem.

### Quick Verification
Run local test/build scripts, check workflow examples and coordination documents.

## Summary and Outlook

Multi-CLI Pilot represents the transformation of AI coding tools from a single CLI to a unified orchestration platform. It solves the fragmentation problem through the provider adapter, improves the quality of collaboration on complex tasks, and expands the tool ecosystem. For teams evaluating or using multiple AI coding agent CLIs, it provides an integration solution, reduces technical debt, and offers a flexible infrastructure for future tool evolution.