About ChengYoung

01

Let AI work locally, on your terms for where data goes.

ChengYoung is a local-first desktop AI agent.

After downloading a model, you can run AI on your own computer — working with code, organizing documents, running workflows and automation tasks. No account required to start; sessions and settings stay on your machine, and the app collects no usage telemetry.

We believe using AI should not require handing over all of your work.

Local-first does not mean everything is offline. Model downloads, software updates, cloud providers you configure yourself, and web access still connect to the corresponding external services. We do our best to make these boundaries explicit, so you know what you are choosing.

Download ChengYoung (free) Read the privacy statement Verify a release

Windows 10/11 (x64) · Free while in development · No usage telemetry

02

Many AI tools can help you work — you should decide where that work happens.

Source code, contracts, internal reports, business data and unpublished writing may all contain information that should not be uploaded casually.

As AI becomes a primary working tool, we do not think users should have to choose between "getting AI capability" and "keeping control of their data".

That is why we built local inference into a desktop application: models run on your own computer, files and sessions stay on your machine — and if you are comfortable with cloud services, you can configure providers yourself and decide deliberately whether to use them.

We will not replace complex technical facts with a single word like "absolutely secure". What we aim for is: less unnecessary data transfer, explicit network behavior, more user control — and a way to verify the promises we publish.

That is the starting point of ChengYoung.

03

Principles we hold on to

Local-first, not cloud-by-default

Once a local inference model is installed and configured, inference can happen entirely on your computer. Sessions and settings are kept in the local data directory and never pass through a ChengYoung server. Subject to hardware and model requirements, local models can run offline.

How to verify: install a model, disconnect from the network, and try a local task; you can also watch the app's network activity with tools like Windows Resource Monitor. See the privacy statement.

No usage telemetry

ChengYoung contains no usage telemetry, device fingerprinting or crash reporting, and collects no behavioral data. This does not mean the app never touches the network: fetching models, checking for updates and calling cloud providers you configured still connect to the corresponding services. The privacy statement separates these scenarios.

Clear boundaries for sensitive actions

Cloud models are configured by you; credentials are stored preferentially in the operating system's credential store (with a configurable file-storage fallback), and gateway tokens live only in the system keychain.

The embedded browser's agent control is off by default. When enabled, it requires per-site authorization; typing and form submission additionally require confirmation for every single action. We want automation to reduce busywork — not to quietly expand its own permissions.

Verifiable releases

Every stable release publishes its artifacts and SHA-256 checksums on GitHub Releases. In-app updates are verified with a minisign signature; a mismatched signature refuses to install.

You should not have to take our word for it. Verify the published checksums yourself.

View GitHub Releases · View the changelog

04

One tool, for real work

ChengYoung aims to be more than a chat window — a desktop tool that takes part in your daily workflows.

Code and technical work

Understand a codebase, analyze problems and discuss changes — let the AI help you get development work done.

Documents and information

Draft, rewrite, summarize and translate; organize reference material and handle everyday documents.

Workflows and automation

Turn repeatable tasks into workflows, combined with skills, MCP extensions and scheduled jobs, to cut down repetitive operations.

Local models and extensions

Run models you have downloaded locally, or connect model services, tools and data sources as you need.

These are scenarios the product supports today — they are not an implication of existing enterprise customers or industry adoption. We will publish real user stories only with genuine feedback and appropriate permission.

05

Origin and attribution

Transparency starts with saying exactly what we publish.

ChengYoung is an independently built distribution of an open-source project: it is published as prebuilt artifacts (Windows installer, checksums, update manifests), provides no source repository, and is not affiliated with or endorsed by the upstream project. The upstream portions are licensed under the Apache License 2.0; ChengYoung's modifications and derivative parts are distributed as prebuilt artifacts and are proprietary software.

The complete origin statement — the upstream project and license, its maintainer, this version's baseline snapshot (version and commit hash), the notice of modifications (Apache-2.0 §4) and the trademark notice — lives on the Provenance page; the bundled LICENSE and NOTICE (including the third-party component list) are readable offline.

One thing to be clear about: the public repository provides prebuilt installers, SHA-256 checksums and update logs only — it does not contain ChengYoung's full source code. ChengYoung is therefore not a product whose complete implementation can be reviewed by anyone. What you can verify is its behavior — whether it connects to the network, where data lands, and what changed between versions.

View provenance View license & third-party notices

06

Who maintains ChengYoung

A project under continuous improvement.

ChengYoung is maintained and published publicly by yunfeng-tian. Development is carried out by a small team covering requirements review, design sign-off, testing and release.

Because the project is distributed without source code and part of the work happens in a private environment, that collaboration does not appear in public commit history. We do not dress this up — judge the product by its actual behavior, not by our one-sided descriptions.

ChengYoung is under active iteration. We focus on making features solid and fixing problems, and on giving every important product decision a clear rationale — rather than making promises for marketing purposes.

If you find a problem, want a capability, or have a better idea, public channels are open.

07

Free today, and about the future

ChengYoung is completely free at this stage.

The current version unlocks all features, with no usage limits.

If a paid tier or cloud service is ever introduced, we will say so on this website in advance — users should never have to guess the business model from usage limits or sudden rule changes.

We will not monetize usage data. The app collects no usage telemetry.

We aim to be transparent about future features and business arrangements. Where things are not yet decided, we prefer to state the current status honestly rather than promise long-term commitments we may not keep.

08

Stay in touch

Make problems visible, and let feedback drive improvement.

• Bug reports and feature requests: GitHub Issues
• Downloads, releases and checksums: GitHub Releases
• Technical release notes: Changelog
• Privacy and data handling: Privacy statement
• Everything else: support@chengyoung.com

We cannot promise every issue will be fixed immediately, but feedback will always have a clear place to land.

If you have a concrete use case or improvement idea, tell us. Real usage problems help us decide what to build next far better than a wishlist detached from reality.