wecode-ai/Wegent

What it solves

Wegent provides a unified AI work system that bridges the gap between local desktop coding and remote cloud-based agent execution. It solves the problem of fragmented AI development workflows where conversations, code changes, and execution environments are isolated across different machines or individuals, allowing teams to share project spaces, automate routine work, and run tasks on the most appropriate device (local, remote, or server-managed).

How it works

Wegent consists of three primary components:

1. Wegent Desktop (Wework): An Electron-based workbench that manages local projects, tasks, and files. It uses the DeepSeek Harness (DSH) runtime to host UI plugins and communicates with an Executor to drive the Codex coding agent.
2. Wegent Backend: A self-hostable core that manages shared project spaces, model configurations, permissions, and remote execution devices.
3. Wegent Web: A browser-based interface for interacting with remote agents that utilize configured models, knowledge, and tools.

Tasks are executed by an Executor layer which drives a concrete coding agent (Codex) to read, modify, and execute commands within a workspace.

Who it’s for

Software developers and engineering teams who want to integrate AI agents into their local development environment while maintaining the ability to scale to remote execution and team collaboration via a self-hosted platform.

Highlights

• Hybrid Execution: Run AI tasks locally on your machine or delegate them to remote work machines and server-managed executors.
• Collaborative Project Spaces: Shared task boards, files, and execution history to keep teams aligned.
• Reusable Agents: Ability to combine specific models, prompts, skills, and knowledge into reusable agent configurations.
• Self-Hostable: Full control over data, permissions, and model configuration by deploying the Web and Backend components via Docker.