Image: GitHub
Why it mattersThe paid always-on AI assistants from xAI, OpenAI and Anthropic only use their own model. An open client that works across chat apps lets a small team keep its own data.
Many people now pay a monthly fee for an always-on AI assistant. xAI charges for Grok Bot. OpenAI charges for dots. Anthropic charges for Claude Duet. A new open-source project does the same job for free, with one limit.
The project is AnyModelBot. The code is on GitHub under an MIT license. The repository was created on 27 October and has 777 stars and 10 forks today. It is ranked 3rd on Trendshift's daily list. Modelflare, a company run by Havenbyte LLC in the US, is the author.
What it does
AnyModelBot is a desktop app for Windows, macOS and Linux. It also runs in Docker on a home server. You create one or more bots. Each bot has a name, a model, a workspace folder and its own memory. You chat with it in the app window. You can also reach the same bot from WeChat, Feishu, Telegram, Slack or Discord.
The bots can read and write files in their folder. They can run commands in a sandbox. They can use your desktop with your approval. Anything outside the folder stops and asks you first. A background task can finish on its own and send the result back to a chat app.
The one limit
Every model goes through Modelflare. Modelflare is the author's own paid API gateway. Every request passes through it, and Modelflare keeps a log of each one. Modelflare says it has 200,000 registered developers. It says one key covers 70-plus models from 13 providers, including GPT, Claude, Gemini, Grok, DeepSeek, Kimi and GLM. Prices are listed per token on modelflare.dev. The app itself has no subscription.
The sign-in opens Modelflare in your browser. For Docker, you paste an API key instead. The project does not document a way to point the app at a different provider. So the code is open, but your model requests still travel through one vendor.
How it compares, in the author's own words
A table in the README places AnyModelBot next to the three paid products. Modelflare says the agent runs on your own computer, the three others run in their vendor's cloud. Modelflare says the models cover GPT, Claude, Gemini, Grok and 70 more. Grok Bot uses xAI models. Dots uses OpenAI. Claude Duet uses Anthropic.
The README notes a limit here. Those details about the three paid products come from public reports as of October 2026. The reports may have changed. Modelflare says AnyModelBot is not affiliated with xAI, OpenAI or Anthropic.
For a team
A small team that already pays for one coding tool can run AnyModelBot as a shared Docker image. Each person gets their own bot in a chat app the team uses. Memory and files stay on the shared machine. The team pays Modelflare per token for what each bot actually does. The audit trail is in two places: the local app, and Modelflare's logs.
The project is six days old on GitHub. The releases page lists installers for every platform. The license permits commercial use.
Source
AnyModelBot on GitHub, by Modelflare.
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
