billmei/every-chatgpt-gui: a curated index of ChatGPT and Claude front-end clients

What this repository actually is

There is no application here. The repository has three top-level entries: .github/, LICENSE and README.md. The README is the product. It opens by saying it exists to "keep track of every graphical user interface alternative to access the API for ChatGPT, Claude, and other LLMs", and it is explicitly modelled on the author's earlier Every Proximity Chat App list. The audience is people who already have an API key and want a chat front end that talks to the API directly rather than through a vendor's own web app. The README also points to a separate guide written for non-technical users, titled "How I pay $1.79 instead of $20/mo for AI using a 5min hack", which is hosted off-repository at billmei.net. That framing tells you the intended reader: someone who wants to pay per token instead of a subscription, and needs a client to do it. Nothing in the repository installs, runs or configures any of the listed clients.

How the list is organised, and what each entry contains

The README is a single flat document divided by deployment shape, not by feature. The visible sections are Open Source, with subsections Web, Browser Extension, Obsidian plugins, Self-Hosted and Desktop. Entries are alphabetical within a category, and the README states that contributors should add apps "under the appropriate category in alphabetical order". Each line follows the same template: a name, then optional links in the form of demo, source, download, or store links for Firefox, Chrome, Edge and Safari. Some entries carry a parenthetical noting a fork, for example SmoothGPT is marked as a fork of ChatGPT UI and WolfeChat as a fork of SmoothGPT. That fork notation is the closest the README comes to editorial comment. There is no description of what any client does, no screenshot, no pricing note, no statement of which model providers each client supports, and no maintenance status. The only signal separating one entry from another is which links it has: a demo URL means you can try it hosted, a source URL means you can read or run the code, and a download URL means a packaged binary exists.

Using the list to pick a client

There is nothing to install from this repository, so the practical workflow is to clone or read the README and then follow links outward. Cloning is useful if you want the list offline or want to diff it over time.

git clone https://github.com/billmei/every-chatgpt-gui.git
cd every-chatgpt-gui

After that, README.md is the only file you need to open. The README gives no build step, no package manager command and no environment variable, because the repository contains no code. If you want to contribute, the README describes the process directly: "If you want to add your app, feel free to open a pull request to add your app to the list." Removal works the same way, by pull request. Practically, the first real use is triage. Open the README, decide whether you want a hosted demo, a browser extension, an Obsidian plugin, a self-hosted server or a desktop binary, and read only that subsection. If you want to try something without running anything, the demo links are the fastest path; if you want to inspect code before trusting it with an API key, the source links are the relevant ones.

The maintenance cost sits in the README, not in code

Because the repository holds only a README and a licence, there is no dependency graph, no build pipeline and no test suite to keep alive. The upgrade cost is editorial. Every entry is a pointer to somebody else's project, and those projects rename themselves, move domains, change licences and go dormant. A list like this decays silently: a demo URL that 404s or a source link that redirects is the failure mode, and nothing in the repository detects it. The README's own contribution rule, add under the appropriate category in alphabetical order, is the entire maintenance protocol. That is cheap to follow and easy to let slip, since alphabetical order is exactly the property that a casual pull request breaks. The last push to the repository was on 2026-08-26, which is recent enough that the list is not stale by default, but recency of the repository does not tell you anything about the state of the projects it links to. Treat each entry as a lead, not as a verified record.

Where the list stops being useful

The README makes no claim about quality, and it should not be read as a ranking. Alphabetical order is the only ordering, so the first entry in a subsection is not the best one. There is no indication of which clients support Claude versus ChatGPT versus other providers, despite the README's opening line naming all three, and no indication of which clients are actively developed. The fork annotations are the sole piece of comparative information present. A reader who wants to know whether a client supports streaming, file upload, conversation export, local model backends or an OpenAI-compatible endpoint will not find it here. The list is also not a security review: it links to browser extensions that request access to pages you visit, and to self-hosted servers that hold your API key, without any note about permissions or key handling. If your decision depends on any of those attributes, this repository is the wrong tool and the linked project's own documentation is the right one. The README is a starting index, and it is silent on everything past the link.

Alternatives to a hand-maintained list

The obvious alternative is a topic-based search on the hosting platform itself, for example browsing repositories tagged with the topics this repository declares: anthropic, chatgpt, claude, gpt-3, gpt-4 and large-language-models. The difference in approach is mechanical rather than editorial. A tag search returns everything matching, including abandoned projects, and gives you metadata such as last commit date and licence on the results page, which this README does not carry. What it does not give you is the human judgement that put an entry in the Web subsection rather than Self-Hosted, and it will not tell you that one entry is a fork of another. A second alternative is to go straight to one of the larger self-hosted clients named in the list and read its documentation, skipping the index entirely. That is faster if you already know you want a server you run yourself; the list is worth the detour only when you have not yet decided on a deployment shape.

Licence and what it means for reuse

The repository is MIT licensed. That covers the README text and any other content the author holds copyright over. It does not cover the projects the README links to, each of which carries its own licence, and it does not cover the linked demo sites or their content. If you copy the list into your own documentation, MIT requires that the copyright notice and permission notice be included; the LICENSE file at the repository root is what you would carry over. If you are evaluating a listed client for commercial use, the licence that matters is that client's, not this one. The README gives no licence information for any entry, so each source link has to be checked individually. Nothing here is legal advice, and the MIT grant says nothing about the trademarks or names of the listed products.

Editorial conclusion

Use this repository if you are starting a search for a ChatGPT UI and want one page that groups web, desktop, self-hosted and extension options instead of a dozen browser tabs. Do not use it if you need install instructions, screenshots, feature matrices or a recommendation, because the README supplies only names, category, and demo and source links. Before adopting any client from it, open that client's own repository and check its last push and licence, since this list carries neither. The specific next step is to read the README's Open Source section, pick one entry, and follow its source link to the project's own install documentation.

Frequently asked questions

Official sources

1. billmei/every-chatgpt-gui on GitHub
2. Issues
3. License: MIT
4. README