Cindy

Cindy is an open-source AI agent from XD Inc. that brings multiple agent harnesses, models and tools into one agent working in your local projects and apps. The client is released under Apache-2.0, with desktop (Electron) and mobile (Expo / React Native) apps in a pnpm monorepo. The homepage lists downloads for Linux, macOS, Windows and Android.

What It Is

Cindy is a local-first AI agent that works with your real files, terminal and signed-in apps. It supports Claude Code, Codex and Pi as harnesses, and models can be mixed across them. You can switch harnesses or models mid-task while workspace, memory, skills and tools stay continuous. Models can come from Cindy's managed service, an existing Coding Plan, your own API key, or local models such as Ollama, vLLM or OpenAI-compatible endpoints.

Multi-Agent Workflows

A single task can be planned, executed in parallel and reviewed by different harness and model combinations, while Cindy holds the task context and acceptance criteria. Work can also be handed to Cindy from Slack or Feishu, run on your computer, and returned to the thread. Scheduled tasks run recurring work and send results to chat, and mobile apps let you check progress and approve sensitive actions.

Control and Safety

Code and file changes start in an isolated workspace, and edits enter a review panel for line-by-line approval. Checkpoints allow rollback. Files, code and chat history stay on your computer by default, and cost and activity are visible in real time. Official builds include usage analytics that the README says do not collect chat or file content.

Extending Cindy

Memory, skills, automation and MCP connections can be customized. A plugin system and public skill marketplace are described as in development. Team and enterprise features, such as SSO, a central AI gateway, credit controls and team skills, are on a waitlist.