modelflare-ai/anymodelbot is an open-source personal AI agent app presented as an alternative to commercial resident AI agents such as Grok Bot, OpenAI dots, and Claude Duet. The latest release announced by the repository is v0.1.21. However, the provided information does not list what changed in this version. It is written in Python and released under the MIT license, and the repository shows 2,696 stars.
Each bot has its own default model, workspace, and memory. It can connect more than 70 models, including GPT, Claude, Gemini, Grok, DeepSeek, Kimi, and GLM, with a Modelflare key, and be used from the app, WeChat, and Feishu. The bots support file operations, command execution, and desktop control, and keep running in the background. Instructions are given for the desktop app, Docker, and running from source. The core is the local-first setup that keeps the user's files, memory, and API keys on the device.
However, model requests are routed through Modelflare to the selected provider, and the service is described as retaining request logs. So "local-first" does not mean that model requests are processed on the device. From the provided information alone, the specific changes in v0.1.21 and the performance of the features cannot be judged; what can be confirmed is a setup that connects multiple models and task paths in a single app.