makecindy/cindy — Consider it done. The open-source AI agent that works out of the box · 想到，就能做到。开源、开箱即用的 AI Agent。

What happened

The makecindy team published the source code for the Cindy client on GitHub. Cindy is an AI agent, which means a program that can carry out tasks for you, not just answer questions. It runs locally on your machine and can use your real files and logged-in apps. The repository holds the desktop app, the mobile app, and shared code packages.

Key facts

• License: Apache-2.0, which lets people view, change, and reuse the code.
• First supported harnesses (the tools that run the AI): Claude Code and Codex. More are planned, and a native harness is in the works.
• Built with Node.js 22.x and pnpm 10.
• Desktop app uses Electron; mobile app uses Expo and React Native.
• The backend service is in a separate repository and is not part of this code.

Some terms explained

A harness is the tool that runs an AI model and lets it act on your computer. A model is the AI brain itself. Cindy lets you mix harnesses and models, and even switch between them in the middle of a task, while your memory, skills, and tools stay the same. You can sign in to the official Cindy service, connect a Claude Code or Codex plan you already pay for, use your own API keys, or run local models.

Why it matters

Open source means anyone can check how the agent works, fix bugs, or add features. Users can teach Cindy to remember corrections, reuse work methods called skills, schedule recurring jobs, and connect company tools through MCP. Because the code is open, people can also see exactly what the agent can access on their computer, which matters for trust and privacy.

What to watch

• Support for more harnesses beyond Claude Code and Codex.
• The native harness, which is still in development.
• Team sharing for skills, which is described as in the making.
• An open plugin marketplace, also still in the making.
• How the project grows through outside contributions under the CONTRIBUTING guide.