The Author's Letter
The story behind the project, its journey, and its author.
1. Why ParrotClaw
I have lived through three eras: the internet, the mobile internet, and artificial intelligence.
Along the way, one feeling has stayed with me: AI is changing the way all software is interacted with. From graphical interfaces that require learning complex operations to an AI mode where you simply talk in natural language and the work gets done — this is not an incremental improvement, but a complete paradigm shift.
What will the world look like 100 years from now? Human productivity will be massively liberated — perhaps as full of imagination as WALL·E.
I hope that in this era, each of us can leave behind something valuable.
ParrotClaw is the thing I want to leave behind.
2. The ParrotClaw journey
The Parrot project goes back to 2023.
2023 — parrot_app
The first versions connected directly to large-model APIs through local configuration, delivering basic AI conversation. The idea was simple: put AI inside an app and make conversation reachable for everyone.
2025 — parrot_desktop
The desktop version introduced Spring AI as its backend, with more complete multi-turn conversation and tool calling. But the architecture was still too heavy — maintenance cost and flexibility were both less than ideal.
2026 — OpenClaw arrives
OpenClaw launched with force this year. I saw its agent orchestration capability — built-in Skills, Plugins, voice interaction, toolchain integration… exactly the backend I had always wanted.
Rather than building an architecture from scratch, why not put the energy into the front-end experience? Build a real AI digital-employee app on top of OpenClaw — that idea appeared naturally.
And so ParrotClaw was born.
How it differs from the official OpenClaw app
ParrotClaw stands on top of OpenClaw — the underlying capabilities are the same as the official app; the difference is that the experience has been made "simple".
The official app is powerful, but it puts up a few barriers for everyday users in China:
- Technical installation — you have to prepare the environment and configure OpenClaw yourself
- Dependence on Google services — most phones in China have no Google Services, so voice recognition does not work
- Developer-oriented interaction — it feels more like a tool than an app that works out of the box
What ParrotClaw improves:
| Aspect | Official OpenClaw app | ParrotClaw |
|---|---|---|
| Getting started | Technical, needs configuration | Simple and easy to use, works out of the box |
| Installation | Install and configure OpenClaw yourself | One-click install on desktop; scan to use on mobile |
| Speech recognition | Relies on Google online services | Offline Chinese ASR (sherpa-onnx) — no network, no Google |
| Speech synthesis | Needs setup, mostly cloud-based | Free TTS, available out of the box |
| Digital human | None | Live2D digital human with lip-synced speech — more fun to talk to |
| Cross-platform | Mobile first | Flutter cross-platform — one experience on desktop and mobile |
In one sentence: ParrotClaw turns "usable" into "easy to use".
ParrotClaw is also open source under the MIT license — the code is fully open, and you are welcome to use and build on it. And like the official app, it supports self-hosting: everything runs on your own devices, so your privacy never leaves your network.
3. A modest contribution
ParrotClaw is only a beginning.
I hope more people will get involved — to suggest ideas for future versions, and if possible, to take part in building them.
One person can walk fast, but a group can walk far.