Cindy AI agent: is the viral "free" Claude Code and Codex app really free?
Short answer
Cindy is an open-source desktop and mobile AI agent that runs Claude Code, Codex and Pi in one app on your own computer, made by XD Inc., the Hong Kong-listed company behind the TapTap game platform. The app is free under Apache-2.0, and you can plug in a Claude, ChatGPT, SuperGrok, Kimi, MiniMax or GitHub Copilot plan you already pay for. What isn't free: Cindy's own model service ($20, $100 or $200 a month in credit). What isn't open: the backend. Usage analytics are on by default, and the "$CINDY" token post came from a brand-new account, not from anything in the repo or on cindy.app.
What is the Cindy AI agent?
Cindy is a client app, not a new model. Per its README, it "brings multiple harnesses, models and tools into one agent" that runs locally "using your real files and logged-in apps." The first supported harnesses are Claude Code and Codex, with Pi added since, and the team says a native harness is in the works.
The pitch is switching without starting over. You can move a task from Claude Code to Codex, or change the model, while the workspace, memory, skills and tools stay put. One task can even be split: the cindy.app example has one agent plan, three workers build in parallel and a fourth review the diff. Cindy can also drive your browser, desktop and phone, take tasks from Slack or Feishu, and run on a schedule.
Under the hood, the desktop app is Electron and the mobile app is Expo/React Native. The README says the real claude-code and codex command-line tools are downloaded per platform when you install, so Cindy is running those tools for you rather than copies of them.
Why is Cindy going viral on X?
On October 26, 2026 at 07:56 UTC (11:56 Dubai), @making_cindy posted: "Cindy made today's GitHub top 1. One open-source agent that brings Claude Code, Codex and other harnesses together to do real work on your desktop and phone. Open source and free." It quoted a GitTrend roundup of the day's fastest-rising repos (2,112,686 views).
When we checked through the X API on the morning of October 27 (Dubai):
- @making_cindy: 2,012,864 views, 2,079 likes, 1,986 bookmarks, 77 reposts and 40 replies. That is about one like per 1,000 views, much lower than most viral AI posts we track (the "Hey Grok" iPhone post had about four). We can't tell from the API whether some of those views were paid promotion.
- Post volume: posts mentioning Cindy with agent, GitHub, Claude Code or Codex went from 78 on October 25 to 398 on October 26 (UTC), plus 107 by 06:27 UTC on October 27.
- GitHub: the repo had 6,829 stars, 819 forks and 2,046 open issues. At 10:30 Dubai on October 27 it was #4 on GitHub Trending with 1,296 stars that day.
Is Cindy AI free?
The software is. The model usage depends on how you connect it. Per the pricing page:
- Free, $0: "All the software, free", including every agent and tool feature, when you bring your own API key or a supported plan. Your provider still bills you as usual.
- Plus, $20 a month: $20 of usage credit for Cindy's managed models. Unused credit rolls over up to $10.
- Super, $100 a month: $100 credit, rollover up to $50.
- Ultra, $200 a month: $200 credit, rollover up to $100.
A monthly plan costs 20% less for the same credit ($16, $80, $160). Pay-as-you-go top-ups start at $10 and don't expire. Cindy also says plainly that "managed plans are not unlimited model access." Managed model families listed include GPT, Claude, Grok, Gemini, DeepSeek, Kimi, GLM, Qwen, MiniMax, Muse, Doubao Seed, Step and Hunyuan, with the exact list shown in the app.
So "open source and free" is accurate for the app. It is the same deal as most agent tools: the expensive part is the model, and you pay for that one way or another.
Can you use your Claude Code or ChatGPT subscription in Cindy?
Cindy says yes. The site lists sign-in with Claude, ChatGPT, SuperGrok, Kimi Code, MiniMax and GitHub Copilot, plus other supported coding plans by key, your own API keys, and local models through Ollama, llama.cpp or LM Studio (local models are marked beta). Your own keys and plans "do not consume Cindy credit", and "your provider's fees and limits still apply." Cindy also notes that "third-party tools may require separate authorization", so check your provider's terms if you rely on a consumer plan for work.
One thing the X post skips: if you want to use Cindy from your phone, the task still runs on your computer. The FAQ says the computer "needs to be online with remote access enabled", and mobile does not move offline work into the cloud.
Get the AI tools worth your time
Era of Minds by Salman Ahmed: one short email with the AI tools and news that actually matter. Free.
You're in. Check your inbox.
Confirm your email and your 50 AI prompts are on their way.
Free. Unsubscribe anytime.
Is Cindy fully open source?
No, only the client. The README is direct about it: "the backend service lives in a separate repository and is not part of this monorepo." The client connects to Cindy's cloud by default, and desktop updates come from Cindy's CDN.
You can choose Skip Sign-In on the login screen to run local agents with no Cindy account, but the README says "server-backed capabilities are unavailable in this state." The Hysen Labs review reaches the same conclusion: don't adopt it if you need a fully self-hosted server. Team and Enterprise plans with a self-hosted gateway are only a waitlist for now.
Who makes Cindy?
XD Inc. Both the repo's NOTICE file and the cindy.app footer say "Copyright 2026 XD Inc." XD is a game developer and publisher listed in Hong Kong since December 2019 (stock code 2400), per its HKEX filings. It runs TapTap, a mobile game community and app store, and publishes games like Ragnarok M: Eternal Love and Sausage Man.
The repo was created on July 22, 2026, and the top contributors by commits are dashhuang (2,247), zqchris (1,732), MagicLizi (1,503) and DavidShenXD (1,112). Release builds come in two flavors, -cn for Mainland China accounts and -global for everyone else, and the README tells developers to pick the right region.
What data does Cindy send?
More than "local-first" might suggest, and the README is honest about it. Release builds include TapDB usage analytics (device identifiers, OS and app version, startup and activity events). Analytics are on by default and "associated with your account ID after sign-in, so this is not fully anonymous." TapDB does not collect chat content, file content or working-directory data.
- Turn it off: on desktop, Settings > About > Usage statistics. On mobile, Usage statistics in Settings.
- Local mode (Skip Sign-In) does not report to TapDB and sends no heartbeats.
- Known mobile gap: the README flags that enterprise SSO sign-in can mark a new install as consented, so turn the switch off yourself.
- Model calls: whatever model you use still receives your prompts and the file content it needs. Cindy's FAQ says "local-first does not mean no data ever leaves your device."
Is there a Cindy crypto token?
Not in anything Cindy publishes that we could find. On October 27, 2026, an X account called @getcindyapp, created at 04:26 UTC that same day and linking to cindy.app, posted at 06:05 UTC: "Cindy is live on @RobinhoodApp" with a crypto contract address. Ninety seconds later it posted "$CINDY has been burned."
The GitHub README, cindy.app and the pricing page say nothing about a token, and none of the @making_cindy posts we read mention one. The app works with API keys, subscriptions and credits; no token is involved. We can't say who runs @getcindyapp, so treat any "Cindy token" post with the same caution as the Oxion DEVWEEK free-model offer: check it against the project's own repo before you send money anywhere.
How do you install Cindy?
- Download from cindy.app or GitHub Releases. The latest stable build, v0.1.115 (October 23), ships macOS Apple silicon and Intel .dmg files (about 260 to 273 MB), a Windows x64 .exe and a Linux x64 .deb. A v0.1.118 beta came out on October 27. The site also lists iOS and Android apps.
- Pick the global build unless you have a Mainland China account.
- Sign in or skip. Connect a Claude, ChatGPT or other plan, add an API key, or choose Skip Sign-In for local agents only.
- Linux: Ubuntu 22.04 and 24.04 install the .deb through apt. Arch and Omarchy use a user-level installer described in the Linux guide.
Building from source needs Node.js 22, pnpm 10 (pnpm 11 isn't supported yet) and Git LFS. One independent test is worth knowing about: MrKeyoor cloned an older commit on October 9, the install took 4,102 MB, and the build failed in a clean Debian container because Electron Forge couldn't find dpkg and fakeroot.
We did not install Cindy or run it ourselves. Everything above comes from the repo, cindy.app and the reviews linked here.
Cindy or Claude Code and Codex on their own?
If you only use one harness and one plan, Cindy adds an app layer, analytics to switch off and a closed backend for its cloud features. Where it makes sense is if you already pay for both Claude and ChatGPT, want to hand a task from one to the other without pasting context, or want to check on a long run from your phone. If shared memory across agents is the part you care about, Beacon takes a different route, capturing sessions from many harnesses into review-gated memory without replacing your tools. And if your worry is the token bill rather than the app, Headroom is the compression proxy people pair with Claude Code.
The pattern is familiar from other viral AI posts we've checked, like the Supabase "12 Claude agents" joke: the headline is true, and the details live in the README.
Get the AI tools worth your time
Era of Minds by Salman Ahmed: one short email with the AI tools and news that actually matter. Free.
You're in. Check your inbox.
Confirm your email and your 50 AI prompts are on their way.
Free. Unsubscribe anytime.
Key facts
- Viral post: @making_cindy, Oct 26, 2026, 07:56 UTC. 2,012,864 views / 2,079 likes / 1,986 bookmarks (checked Oct 27).
- Repo: makecindy/cindy, Apache-2.0, TypeScript, created Jul 22, 2026. 6,829 stars, 819 forks, 2,046 open issues (Oct 27).
- Harnesses: Claude Code, Codex and Pi. Switch harness or model mid-task; memory, skills and tools carry over.
- Free with your own key or plan. Managed credit: Plus $20, Super $100, Ultra $200 a month ($16 / $80 / $160 on a monthly plan).
- Backend is closed and in a separate repo. Skip Sign-In runs local agents only.
- TapDB analytics on by default, tied to account ID; no chat or file content. Off in Settings > About > Usage statistics.
- Made by XD Inc. (HKEX 2400, TapTap). Separate -cn and -global builds.
- $CINDY token post came from @getcindyapp, an account created Oct 27; no token in the README or on cindy.app.
FAQ
What is the Cindy AI agent?
Cindy is a desktop and mobile app that runs AI coding agents on your own computer. It puts Claude Code, Codex and Pi in one workspace and lets you switch harness or model in the middle of a task while memory, skills and tools carry over. The client is open source under Apache-2.0 at github.com/makecindy/cindy, and the company behind it is XD Inc.
Is Cindy AI free?
The app is free. If you connect your own API key or a supported subscription (Claude, ChatGPT, SuperGrok, Kimi Code, MiniMax or GitHub Copilot), Cindy charges no software fee, but your provider's fees and limits still apply. Cindy's own managed model service is paid: Plus $20, Super $100 and Ultra $200 a month in usage credit, or $16, $80 and $160 on a monthly plan.
Can I use my Claude Code or ChatGPT subscription in Cindy?
Yes, according to cindy.app. You can sign in with Claude, ChatGPT, SuperGrok, Kimi Code, MiniMax or GitHub Copilot, or connect other supported coding plans with a key. Usage goes against that plan's own limits and does not use Cindy credit. The desktop app downloads the real Claude Code and Codex command-line tools during install and runs them for you.
Is Cindy fully open source?
No. Only the client is. The README says the GitHub repo holds the desktop and mobile apps and shared packages, and that the backend service lives in a separate repository that is not part of the open-source code. You can skip sign-in and run local agents without an account, but server-backed features are unavailable in that mode.
Who makes Cindy?
XD Inc. The repo's NOTICE file and the cindy.app footer both say Copyright 2026 XD Inc. XD Inc. is a Hong Kong-listed game company (stock code 2400) that runs the TapTap game platform and publishes games like Ragnarok M: Eternal Love and Sausage Man. Release builds come in separate versions for Mainland China and global accounts.
Does Cindy collect data?
Release builds include TapDB usage analytics, which are on by default and linked to your account ID after sign-in. The README says TapDB does not collect chat content, file content or working-directory data. You can turn it off on desktop in Settings, About, Usage statistics. Local mode without sign-in does not report to TapDB. Model requests still send your prompts and file context to the model provider you pick.
Is there a Cindy crypto token?
Not one we could find in Cindy's own materials. On October 27, 2026, a new X account, @getcindyapp, created that day, posted that Cindy is live on Robinhood with a contract address, then posted that $CINDY has been burned. The GitHub README, cindy.app and its pricing page say nothing about any token, and you do not need one to use the app. Treat token posts with care.
Sources
- @making_cindy on X, October 26, 2026
- @GitTrend0x on X, October 25, 2026
- GitHub: makecindy/cindy README
- GitHub: makecindy/cindy releases
- GitHub: Cindy Linux installation guide
- GitHub: Cindy NOTICE file
- cindy.app home page and FAQ
- cindy.app pricing
- Hysen Labs: Cindy AI agent client review
- MrKeyoor: Cindy build test
- XD Inc.: About us
- XD Inc. annual results announcement, HKEX, March 2025
- @getcindyapp on X, October 27, 2026
Keep reading
Get AI news decoded weekly.
Decoded is a free weekly email from Era of Minds. The AI stories that actually matter, in plain English. Five minutes, no hype.
You're in. Check your inbox.
Confirm your email and your 50 AI prompts are on their way.
Free. Unsubscribe anytime. Free gift: 50 AI prompts that actually work, sent the moment you join. More about Decoded