Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseCode / SDK · ActionableGitHub Trending AI→ original

Open-source agent Cindy gathered 1,446 GitHub stars in a day and combined Claude Code with Codex

Cindy is an open-source AI agent that unifies a work computer, a home PC, and a phone into a single environment. You can start a session on a laptop and continue it from your phone, and model subscriptions set up on one computer are available on another. Claude Code and Codex run inside it, the client is released under Apache-2.0, and the project collected 1,446 GitHub stars in a day. The backend lives in a separate repository, so cloud features still require a Cindy account.

AI-processed from GitHub Trending AI; edited by Hamidun News
Open-source agent Cindy gathered 1,446 GitHub stars in a day and combined Claude Code with Codex
Source: GitHub Trending AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceGitHub Trending AI ↗
Editor-in-Chief
Client under Apache-2.0macOS, Windows, Linux, iOS, and Android
Key metric #2The "Skip Sign-In" mode runs local agents without an account
Key metric #3Usage statistics are enabled by default and can be turned off in settings

Plenty of agents for code and work tasks already exist, and most developers end up with a zoo: Claude Code on one laptop, Codex on another, subscriptions and API keys scattered across machines, and no way to pick up a conversation from a phone. The Cindy project, from the makecindy/cindy repository, tries to bring all of this into a single shell. It topped GitHub Trending and gained 1,446 stars in a day. Here is what's inside and where the catch is.

One session across all devices

The core idea: a work computer, a home desktop, a laptop, and a phone operate as one "companion." You can open a session from another device, continue the conversation, and access files without copying everything into a new chat. This is not full disk synchronization: project files stay on the computer where they live.

The phone app also includes remote desktop. If a chat isn't enough, you can open programs, type text, and control the connected computer. The conditions are simple: the machine must be powered on, you must be signed in with the same account, and the necessary permissions must be granted separately for remote access and for the desktop.

Models can also be shared between devices. A provider configured on one computer will be available on another, including subscriptions, API keys, and local models. The authors stress that the computer becomes your server, so there is no need to rent or set up a separate server.

Claude Code, Codex, and agent teams

Cindy doesn't replace existing agents; it wraps them. Claude Code and Codex plug in as "engines," while the shared interface, multi-device access, and tools stay with Cindy. There is also a built-in lightweight, extensible agent that supports multiple model providers. You can change the combination of engine and model mid-task. But the README includes a candid caveat: the session itself stays in Cindy, and when you switch engines, context is passed along or rebuilt, but the agents' internal states do not become identical.

For large tasks, the project offers a team mode: different agents plan, write code in parallel, and independently verify the result toward a shared goal. The repository mentions orchestration documentation for multiple agents called Orca.

Beyond that, the features are familiar from agent platforms. You can create your own companions with a role, instructions, and recurring tasks, such as a reviewer or a researcher. Useful techniques can be saved to memory and turned into Skills, which can be shared through SkillHub. Tasks can also be assigned through chat channels like Slack and Telegram.

What is open and what isn't

The desktop and mobile client is distributed under Apache-2.0. It is a pnpm monorepo: a mobile app built on Expo and React Native, plus shared packages for authentication, device connectivity, agent orchestration, and model providers. The desktop build ships with claude-code, codex, ripgrep, and pi, which are downloaded for the target platform at build time. macOS, Windows, Linux, iOS, and Android are supported.

The server side, however, lives in a separate repository and is not part of the monorepo. By default, the client connects to Cindy's official cloud. External developers are advised not to deploy their own server and instead sign in with their own account in the dev build. Without an account, there is a "Skip Sign-In" mode: local agents work, but features that depend on the server are unavailable. Cloud companions sync conversations, memory, skills, and attachments through Cindy's services.

The analytics section deserves a close read. Launch and return statistics in official builds are enabled by default and include a device identifier and OS and app versions. After sign-in, daily model selection and total token counts are added. According to the authors, chats, prompts, files, and working directories are not included. You can disable it in settings. The README also describes an edge case: with corporate SSO, migration may record consent even on a new installation while statistics remain enabled. Separately, heartbeat signals run during cloud sign-in, and diagnostic logs are uploaded manually.

Practical takeaway

Cindy is of interest to people who already pay for Claude Code or Codex and want one session across devices without building their own server. Before rolling it out on a team, check three things: which data flows through Cindy's cloud, whether analytics is enabled, and whether the client license is sufficient if the backend remains separate. A free client does not mean free tokens: provider rate limits and pricing still apply. The Apache-2.0 license does not automatically extend to model weights, datasets, prompts, or trademarks.

Story

Key Questions Answered

What are the key details on One session across all devices?

The core idea: a work computer, a home desktop, a laptop, and a phone operate as one "companion." You can open a session from another device, continue the conversation, and access files without copying everything into a new chat. This is not full disk synchronization: project files stay on the computer where they live. The phone app also includes remote desktop. If a chat isn't enough, you can open programs, type text, and control the connected computer.

What are the key details on Claude Code, Codex, and agent teams?

Cindy doesn't replace existing agents; it wraps them. Claude Code and Codex plug in as "engines," while the shared interface, multi-device access, and tools stay with Cindy. There is also a built-in lightweight, extensible agent that supports multiple model providers. You can change the combination of engine and model mid-task.

What is open and what isn't?

The desktop and mobile client is distributed under Apache-2.0. It is a pnpm monorepo: a mobile app built on Expo and React Native, plus shared packages for authentication, device connectivity, agent orchestration, and model providers. The desktop build ships with claude-code, codex, ripgrep, and pi, which are downloaded for the target platform at build time. macOS, Windows, Linux, iOS, and Android are supported. The server side, however, lives in a separate repository and is not part of the monorepo.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Hands-On AI Engineering · Hamidun Academy

Build & Ship Production AI Agents Without Boilerplate Code

Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.

Start 72h Free Trial →Vibe-Coding Intensive
What do you think?
Loading comments…