Openwork provides a desktop interface for deepagentsjs, an opinionated harness for building AI agents with filesystem capabilities, planning, and subagent delegation. It targets developers building sophisticated AI agents that require direct interaction with the local filesystem and shell environment, offering a structured way to manage complex agent workflows.

How It Works

Openwork serves as a desktop graphical interface built upon deepagentsjs, an opinionated framework for developing advanced AI agents. Its core design enables these agents to interact directly with the host system's filesystem and execute arbitrary shell commands. This direct access, coupled with capabilities for planning and delegating tasks to subagents, allows for sophisticated agent workflows that can manage files, run scripts, and orchestrate complex operations. The advantage lies in providing a user-facing environment for managing these powerful, potentially risky, agent actions.

Quick Start & Requirements

• Installation:
  • Directly via npx: npx openwork
  • Globally via npm: npm install -g openwork
  • From source: git clone https://github.com/langchain-ai/openwork.git, cd openwork, npm install, npm run dev
• Prerequisites: Node.js 18+.
• Supported Models: Anthropic (Claude 4.5/4.1), OpenAI (GPT-5.2/5.1/o3/o4/o1), Google (Gemini 3/2.5 Pro/Flash).
• Configuration: Models can be configured in-app via the settings panel.

Highlighted Details

• Direct System Integration: Provides AI agents with direct read/write access to the local filesystem.
• Shell Command Execution: Enables agents to execute arbitrary commands within the host operating system's shell.
• Advanced Agent Architecture: Supports planning capabilities and delegation to subagents for complex, multi-step tasks.
• Broad LLM Support: Integrates with leading AI models from Anthropic (Claude), OpenAI (GPT), and Google (Gemini).

Maintenance & Community

Bug reporting is handled via GitHub Issues. Contribution guidelines are available in CONTRIBUTING.md.

Licensing & Compatibility

The project is licensed under the MIT License. This license is generally permissive for commercial use and closed-source linking.

Limitations & Caveats

[CAUTION] The project grants AI agents direct access to the filesystem and the ability to execute shell commands. Users must exercise extreme caution, review tool calls before approval, and only run in trusted workspaces due to significant security implications.