ChatALL: run one prompt across ChatGPT, Gemini, Claude and 30+ bots at once

The problem ChatALL solves for prompt comparers

LLM behaviour is inconsistent. The README states this plainly: bots "can be random and different bots excel at different tasks", and it advises against trying them one by one. ChatALL, whose Chinese name is 齐叨, sends one prompt to several bots concurrently so the answers arrive together and can be judged against each other.

The README names three audiences. Gurus of LLMs who want the best answer or creation. Researchers who want to compare strengths and weaknesses across fields. Developers of LLM applications who want to debug prompts and pick a foundation model. All three share one workflow: the same input, many outputs, a human deciding which is better. That is a comparison tool, not a chatbot. If you only ever talk to one model, the concurrency buys you nothing and the setup cost is pure overhead.

The supported list is broad and geographically mixed: ChatGPT, Gemini, Claude, Copilot, Perplexity, Phind, Poe, You, Pi, Kimi, ChatGLM2, MOSS, Baidu ERNIE, iFLYTEK SPARK, Tongyi Qianwen, SkyWork, Vicuna, Llama 2, Falcon 180B, WizardLM, Zephyr, Gemma, Code Llama, Groq Cloud, Cohere, xAI Grok and more. The README says more are coming and points readers to issues where they can upvote their favourite bots.

Web sessions versus API keys: the two connection paths

The supported-bots table has two columns that matter more than the model names: Web Access and API. Some bots are reachable only through the web interface, some only through an API, and several support both. ChatGPT, for example, lists Yes in both columns and the notes mention web browsing and Azure OpenAI service. Claude and Gemini also list both. Baidu ERNIE, Cohere Aya 23, Cohere Command R, Groq Cloud and xAI Grok list API only. Character.AI, Perplexity, Phind, Poe, YouChat, ChatGLM2, MOSS, Vicuna, Llama 2, Falcon 180B, WizardLM, Zephyr, Gemma, Code Llama, Gradio and HuggingChat list web access only.

That split defines the operating model. Web-only bots are driven through the page, which means you log in inside the app and the session is what carries your request. The README marks ChatGLM2 and Vicuna as requiring no login, which is the exception. API bots need a key you supply. So the app is a fan-out layer: one input, N adapters, N responses collected and displayed together.

The table also carries honest gaps. iFLYTEK SPARK shows web access yes and API "Coming soon". Poe, SkyWork and Tongyi Qianwen show API "Coming soon". Dedao Learning Assistant is listed as "Coming soon" with no API. Those entries tell you the maintainers track availability per bot rather than promising uniform coverage.

The stack behind this is a Vue 3 and Vuetify 3 front end inside Electron, with Dexie for local storage and LangChain integration packages for several providers. The package.json depends on @langchain/openai, @langchain/anthropic, @langchain/google-genai, @langchain/cohere, @langchain/groq and @langchain/xai, which lines up with the API-only entries in the table. Conversations live locally through Dexie, and dexie-export-import is present, so history is a local database rather than a server-side account.

Installing ChatALL and sending your first concurrent prompt

The README does not walk through a build. It says the whole onboarding is to "download, install and ask", and links to the GitHub releases page. For most users that is the path: pick the installer for Windows, macOS or Linux from the release assets and run it.

If you want to run from source, the repository is a Node project. The package.json sets engines.node to ^20, so use Node 20 before anything else. Clone the repository, install dependencies, then start the Electron app in development mode.

git clone https://github.com/ai-shifu/ChatALL.git
cd ChatALL
npm install
npm run electron:serve

The postinstall script runs electron-builder install-app-deps, which rebuilds native dependencies for Electron; expect that step to take a while on a cold cache. npm run electron:serve launches the desktop window rather than a browser tab. There is also npm run serve for the plain Vue dev server, and npm run electron:build to produce a packaged app.

Once the window is open, the flow described by the README is: select the bots you want in the conversation, type one prompt, and send. Each selected bot receives the same prompt and the responses stream back in parallel into the same view. Web-only bots will ask you to log in the first time, because the session is what authenticates the request. API-only bots need a key entered in the app's settings before they will answer. The README's own framing of the result is that you "discover the best" answer by looking at all of them at once.

Where ChatALL breaks: session drift and unsupported bots

The weak point is the web-session path. A large share of the table is web access only, and web interfaces change without notice. When a provider reshuffles its page or adds a check, the adapter for that bot stops working until someone updates it. The release cadence in the repository reflects this: v1.85.110 landed on 2025-05-22, v1.84.109 on 2025-03-16, and v1.83.108 on 2024-10-08. The last push to the repository was on 2026-09-10, so work continues, but a fix for a broken bot depends on that bot being reported and patched.

API bots have the opposite problem: they are stable but metered. You pay per token through your own key, and a wide fan-out multiplies that cost by the number of bots you select. Sending one prompt to eight API bots is eight billed calls, which is easy to forget when the send button looks like a single action.

There is also a scope limit. ChatALL is an Electron desktop application, not a server. There is no documented headless mode, no HTTP endpoint for other programs to call, and no multi-user account model; conversations sit in a local Dexie database. If your goal is to put multi-model comparison behind an internal API for a team, this is the wrong shape of tool. The README's audience list does not include service operators either.

Finally, the README is silent on rollback, on what happens to stored conversations when you upgrade, and on rate-limit behaviour when many bots are queried at once. Treat those as unverified rather than solved.

ChatALL versus ChatHub and Poe

The natural comparisons are ChatHub and Poe, both of which appear in what people search for around this project. The difference is in where the comparison happens.

Poe is a hosted product. You use it through an account on someone else's infrastructure, and the set of models is whatever that service offers. ChatALL is the opposite: a local Electron app you install, with no account layer of its own, pulling from bots you already have access to. Your prompts and history stay in a local database. The trade is that you maintain the connections, and a broken adapter is your problem until it is fixed upstream.

ChatHub is the closer alternative because it is also a browser-side multi-chat interface. The practical difference is packaging and reach. ChatALL ships as a desktop app for Windows, macOS and Linux with installers on its releases page, and its bot list leans heavily toward Chinese providers such as Baidu ERNIE, iFLYTEK SPARK, Tongyi Qianwen, ChatGLM2, MOSS, Kimi and SkyWork, alongside the Western ones. If your work involves comparing Chinese and Western models in one view, that coverage is the reason to pick ChatALL over a browser extension. If you want zero installation and are happy with a hosted account, Poe is the shorter path.

None of these tools is a substitute for an evaluation harness. They are for a human reading answers side by side, not for scoring outputs at scale.

Licence, maintenance and what an upgrade costs you

ChatALL is licensed under Apache-2.0. That is a permissive licence, and the repository ships the LICENSE file at the top level. Apache-2.0 includes an explicit patent grant and requires that you keep the licence and notices when you redistribute. It does not give you the models: each bot you connect to is governed by its own provider's terms, and using a web session to drive a bot is a different relationship from using that provider's API. That distinction is worth checking against each provider's terms before you build anything commercial on top. This is not legal advice; read the LICENSE file and the providers' terms yourself.

On maintenance, the repository is not archived and the last push was on 2026-09-10. The most recent tagged release in the list is v1.85.110 from 2025-05-22, so tagged releases lag the commit activity. For an app whose value depends on scraping-friendly web adapters, that lag is the number to watch: a tag from May 2025 against a repository still receiving commits in September 2026 suggests fixes may land on main before they reach a release build. If a bot you depend on breaks, check main before assuming the project is dead.

Upgrade cost is mostly re-authentication. Because web-only bots rely on sessions, a new build can require logging in again, and the README does not document whether conversation history survives an upgrade. The app stores data locally via Dexie, and dexie-export-import is a dependency, so exporting before a major version jump is the cautious move. Verify that yourself rather than trusting the absence of a warning.

Editorial conclusion

Adopt ChatALL if you compare model outputs or debug prompts and want one window instead of a tab bar. Skip it if you need a headless, scriptable, multi-tenant service: it is an Electron desktop app, and many bots are driven through web sessions rather than APIs. Before relying on it, verify that your target bots still connect, since the README lists several as web access only and a few as coming soon, and check the release page for the build matching your platform.

Frequently asked questions

Official sources

1. ai-shifu/ChatALL on GitHub
2. License: Apache-2.0
3. Project website
4. README
5. Releases