Cindy Operational

Cindy is an open-source AI Agent workspace that runs on your own computer. It integrates multiple Agent engines, including Claude Code, Codex, and Pi, so you can switch between them at any time without losing context. It supports cross-device collaboration across macOS/Windows/Linux/iOS/Android, can work directly with your browser sessions and files, and applies isolated review before changes are finalized. Cindy supports MCP, Skill extensions, and scheduled tasks. The client is open source under Apache-2.0.

Follow the WeChat account and reply “开源” to join the AI open-source project community.
Multi-engine integration: Bring Claude Code, Codex, Pi, and other Agents into one workspace. Switch engines at any time while keeping tasks and memory intact.
Cross-device collaboration: Connect your computer, phone, and tablet. Continue the same task on another device and share models configured on other devices.
Remote desktop: Remotely control a connected computer from your phone, open applications, and run commands without a VPN or intranet tunneling.
Real execution: Directly operate local files, the command line, and logged-in browsers. Work in an isolated workspace first, then apply changes only after review approval.
Review mechanism: Every change appears in the review panel for individual approval. Revert to the previous checkpoint if needed.
Multi-Agent teamwork: Have multiple Agents divide planning, implement in parallel, and conduct independent reviews while sharing the same goal.
Skill and MCP extensions: Use the MCP protocol to connect community tools, and install, share, and reuse workflows through the SkillHub marketplace.
Persistent memory: Cindy remembers corrections automatically, making preferences and knowledge increasingly convenient to use.
Scheduled tasks: Supports automated workflows such as daily summaries and regular checks.

Download and install: Visit the Cindy website https://cindy.app/ or GitHub Releases, and choose the installer for your system.
Create a task: Click “New Task” and describe your needs specifically, such as “Write a Python script to organize the photos on drive D by shooting date.”
Monitor execution: Cindy works in an isolated workspace, while token usage and progress are visible in real time in the interface.
Review and approve: Check changes individually in the review panel. Changes are applied only after confirmation, and you can revert and start over if needed.
Extend capabilities (optional): Install Skills from SkillHub, connect an MCP Server, or connect Feishu/Slack for task assignment through IM.
Mobile integration (optional): Install the mobile app to remotely view task progress, approve actions, or use your phone as a computer remote control.

Multi-engine integration: Claude Code, Codex, and Pi coexist in one workspace, allowing you to switch engines without losing context.
Full cross-platform support: Seamlessly continue tasks between desktop and mobile. Configure models once and use them everywhere.
Real execution: Directly operate local files, the terminal, and logged-in browsers instead of only producing text suggestions.
Secure review mechanism: Changes run in an isolated area first and are applied only after individual approval. They can be reverted and audited.
Zero server costs: Your computer is the server. Cross-device collaboration requires no rented server or intranet tunneling setup.
Gets smarter with use: Memory and Skills turn corrections and workflows into reusable personal assets.

Project website: https://cindy.app/
GitHub repository: https://github.com/makecindy/cindy

Continue development across devices: Assign a refactoring task on your work computer, review progress and approve changes on your phone during the commute, then verify it when you get home.
Browser-session automation: Use logged-in GitLab, Confluence, and Feishu accounts to organize documents and synchronize data without sharing account passwords.
Automated team daily reports: Scheduled tasks collect GitHub Trending and project updates, summarize them using a fixed template, and send them to Feishu or Slack groups.
Multi-Agent collaborative development: Plan with Claude Code, implement with Codex, and use an independent Agent for review while sharing acceptance criteria to handle complex projects.
Personal knowledge management: Teach Cindy a workflow such as “organize this week's materials and write a weekly report,” save it as a Skill, and trigger it with one sentence in the future. It becomes increasingly aligned with your personal habits.