English · 简体中文 · Documentation
SoloQueue is an open-source, local-first AI workspace for personal use. Work with it around your projects to keep conversations going and move tasks forward. It can use files, commands, and external tools to get work done; extend its capabilities with Skills; build long-term memory; and coordinate multiple AI agents on complex tasks.
SoloQueue draws inspiration from OpenClaw and my exploration of multi-agent orchestration and collaboration. After studying and practicing with a range of Agent Harness approaches, I built SoloQueue's core in Go and optimized it for DeepSeek, aiming to balance execution performance, cost, and everyday usability. It is also an ongoing engineering practice in building an Agent Harness.
SoloQueue has been running on my personal computer for a long time. It is lightweight and easy to deploy and extend. Connect a compatible model provider to run it on a personal computer or server, then use the web interface, QQ, WeChat, or Telegram to start tasks, track progress, and receive results.
Building from source requires Go 1.25.8, Node.js, pnpm, and Make:
git clone https://github.com/Fize/soloQueue.git
cd soloQueue
make build
./soloqueue startOpen http://127.0.0.1:57689, configure a model, and add a project directory to get started. You can enter the provider API key in the Web Console.
For other model providers, Docker deployment, and troubleshooting, see the Getting Started guide.
SoloQueue listens on 127.0.0.1 by default. Its tools inherit the system permissions of the SoloQueue process and can read files and run commands. For stronger isolation, running SoloQueue in a container or virtual machine is strongly recommended.