cindy: AI Agent for Your Computers and Phone

Overview

Cindy is an open-source AI agent for macOS, Windows, Linux, iOS, and Android, built around the Claude Code and Codex coding harnesses and the extensible, open-source Pi agent. It brings those agents into a shared interface and connects your computers and phone, so you can continue work across devices and operate a connected computer remotely. Your computer acts as the server, rather than requiring you to rent or self-host one.

It is aimed at people who want an agent to do practical work with their files, browser, and apps, and who want to shape a companion with skills, memory, and automations. You can use supported subscriptions, API keys, local models, or Cindy's managed model service. This repository contains the desktop and mobile clients, not the backend service.

Key Features

• Connect computers and phones to continue sessions and access files on another device. Project files remain on their own computer rather than syncing across the whole disk.
• Use Claude Code, Codex, or Pi through a shared interface, and choose supported harness and model combinations.
• Operate a connected computer from a phone using remote desktop, with the required permissions enabled.
• Coordinate agents for planning, parallel implementation, and independent review.
• Create companions with their own roles, instructions, skills, and recurring work.
• Extend workflows with Skills, plugins, MCP, connected chat channels, and automations.
• Work with outputs that can include charts, interactive controls, and rich HTML.

Use Cases

• A developer using Claude Code or Codex can keep those workflows alongside Pi and continue a session from another connected device.
• Someone moving between a work computer and a personal laptop can access sessions and project files on the computer where the files live.
• A phone user who needs to interact with an app on a connected desktop can use remote desktop instead of relying on chat alone.
• A team or individual can use separate agents for planning, implementation, and review, then build recurring routines around the work.
• A user can create a research or review companion and extend it with reusable Skills or connected tools.

Getting Started

For a minimal development checkout:

git clone https://github.com/makecindy/cindy.git
cd cindy
git lfs pull
pnpm install

The README lists Node.js 22.x, pnpm 10.x, and Git LFS as prerequisites. See the README and contributor guide for setup and development details. For end-user builds, use the Cindy download page.

Alternatives

• DuoDesk: DuoDesk links devices for screen, app, file, and input sharing, while Cindy connects devices for cross-device work with AI coding agents.

Considerations

• Cindy is less than six months old and changes quickly. Its recent activity includes frequent commits and releases, but its age makes long-term behavior and compatibility less established.
• The measured contributor base is broad rather than dependent mostly on one person. Recent issue closures and pull-request merges indicate active maintenance, though the available figures do not establish how quickly every report is handled.
• The client is Apache-2.0 licensed, but the backend is outside this repository. Hosted features require a Cindy account, and model providers or Cindy's service can have separate billing and limits.
• The client connects to Cindy's official cloud services by default. Skip Sign-In supports local agents, but server-backed capabilities are unavailable in that mode.
• Official distribution builds include usage analytics, which can be disabled in settings. The README documents a Mobile enterprise SSO consent edge case and distinguishes analytics from heartbeats and diagnostic logs.
• The repository's measured requirements list Node.js 22.12 or newer and CI, but no test suite; the README additionally calls for pnpm 10.x and Git LFS for development.