# PilotDeck — Task-Oriented AI Agent Productivity Platform

> An open-source AI agent productivity platform by OpenBMB that orchestrates task-oriented agents with a visual workspace, structured planning, and multi-model support.

## Install

Save in your project root:

# PilotDeck — Task-Oriented AI Agent Productivity Platform

## Quick Use
```bash
git clone https://github.com/OpenBMB/PilotDeck.git
cd PilotDeck
npm install
npm run dev
# Open http://localhost:3000 to access the workspace
```

## Introduction
PilotDeck is an open-source platform by OpenBMB (the research group behind ChatDev and XAgent) for managing task-oriented AI agents through a visual workspace. It provides a structured environment where agents break down complex tasks into plannable steps, execute them with tool access, and report results in an organized dashboard. PilotDeck supports multiple LLM backends and is designed for both individual developers and teams.

## What PilotDeck Does
- Provides a visual workspace for creating, managing, and monitoring AI agent tasks
- Breaks complex tasks into structured plans with step-by-step execution
- Supports multiple LLM providers as agent backends simultaneously
- Offers real-time progress tracking with detailed execution logs
- Enables team collaboration on agent-managed projects

## Architecture Overview
PilotDeck is a TypeScript web application with a React frontend and a Node.js backend. The planning engine decomposes user tasks into directed acyclic graphs of subtasks, each assigned to an agent with appropriate tools. The execution runtime manages agent lifecycles, tool permissions, and result aggregation. A WebSocket layer provides real-time updates to the dashboard. State is persisted locally with optional database backends for team deployments.

## Self-Hosting & Configuration
- Clone the repository and install dependencies with npm
- Configure LLM provider API keys in the environment file
- Start the development server for local use
- Deploy with Docker for team or production environments
- Customize agent tool access through the admin settings panel

## Key Features
- Visual task planning with automatic subtask decomposition
- Multi-model support lets different agents use different LLMs
- Real-time execution dashboard with step-by-step progress
- Tool permission system controls what agents can access
- Exportable task reports for documentation and review

## Comparison with Similar Tools
- **AutoGen** — code-first multi-agent framework; PilotDeck adds a visual management layer
- **CrewAI** — role-based agent orchestration in Python; PilotDeck is a web-based workspace
- **Dify** — LLMOps platform focused on chatbots; PilotDeck targets task execution workflows
- **Plane** — project management for humans; PilotDeck manages AI agent tasks
- **n8n** — workflow automation via nodes; PilotDeck is purpose-built for agent task orchestration

## FAQ
**Q: Which LLM providers are supported?**
A: OpenAI, Anthropic, Google, and any OpenAI-compatible endpoint including local models.

**Q: Can agents use external tools?**
A: Yes, agents can access web search, code execution, file operations, and custom MCP tools.

**Q: Is it suitable for production use?**
A: PilotDeck is actively developed and suitable for individual and small team use, with enterprise features in progress.

**Q: Does it support Chinese language interfaces?**
A: Yes, PilotDeck supports both English and Chinese interfaces, reflecting OpenBMB's research origins.

## Sources
- https://github.com/OpenBMB/PilotDeck
- https://github.com/OpenBMB/PilotDeck/blob/main/README.md

---
Source: https://tokrepo.com/en/workflows/asset-9b1f6760
Author: AI Open Source