Cindy
Cindy is an open-source AI agent that works out of the box, bringing multiple harnesses, models, and tools into one agent to finish real work in your projects and apps. Ready from day one and yours to shape over time.
Visit makecindy/cindyOverview
Cindy is an open-source AI agent and local-first client for desktop and mobile. She runs on your own machine with your real files and logged-in apps, supports multiple harnesses such as Claude Code and Codex, and lets supported models run across supported harnesses. Harnesses and models can switch mid-task while workspace, memory, skills, and tools stay continuous. The repository contains the open-source client, including desktop and mobile apps and shared packages; the backend service lives in a separate repository.
Key Features
- Multiple agent harnesses: Claude Code, Codex, and Pi are supported, with more harnesses on the way and a native Cindy harness in development.
- Harness and model routing: every supported model runs across every supported harness, and users can switch harnesses or models mid-task without restarting the work.
- Multi-agent orchestration: a task can be planned, executed in parallel by workers, and independently reviewed using different harness and model combinations while Cindy holds context and acceptance criteria.
- Ready-to-work model access: use the managed Cindy service, an existing Coding Plan, your own API key, or local models, with 20+ tools and sensible defaults.
- Local workspace and integrations: work with projects, files, apps, terminal, Slack, Feishu, GitHub, Jira, SSH, servers, Android, MCP, skills, and voice.
- Control and safety: ask-first behavior at critical moments, isolated workspaces, line-by-line review, rollback checkpoints, local workspace retention, and visible work and cost.
Use Cases
- Individual developers delegating coding and project tasks on their own computer while reusing existing model subscriptions or API keys.
- Teams using Slack or Feishu to @ Cindy, run tasks on a member's computer, return results to the thread, and schedule recurring work.
- Advanced multi-agent workflows that plan, execute in parallel, and review work across different harness and model combinations.
- Team and enterprise administration with centralized model access, credit and daily limits, distributed team skills, SSO, and optional self-hosted Gateway.
Getting Started
- Install Node.js 22.x, pnpm 10.x, and Git LFS.
- Clone the repository, run git lfs pull, and then run pnpm install.
- Contributor setup, validation, and submission workflow are documented in CONTRIBUTING.en.md.
- For remote development, use pnpm restart:desktop:remote --region=cn for Mainland China Cindy accounts or --region=global for global accounts.
Deployment & Requirements
- Node.js 22.x, pnpm 10.x (v11 is not yet supported), and Git LFS are required for the repository.
- The client is organized as a pnpm monorepo with an Electron desktop app, an Expo/React Native mobile client, and shared packages; the backend service is in a separate repository.
- Official distribution builds connect to Cindy's official cloud services by default; desktop auto-updates also come from the official CDN.
- Linux users have an installation guide for Ubuntu, Arch Linux, and Omarchy covering installation, updates, keyring setup, and migration from a manual install.
Before You Adopt
- License: Apache-2.0. Review its terms before using, modifying, or distributing the project.
- Some advertised capabilities are still in development, including the public marketplace and plugin system.
- Proactive briefs and work suggestions are not yet available and are being developed.
- Using Skip Sign-In runs local agents without a Cindy account, but server-backed capabilities are unavailable in that state.