AnyModelBot serves as a local-first AI agent application, designed for users seeking persistent and customizable assistants on their personal Windows, macOS, or Linux machines. This innovative tool enables individuals to craft bots featuring distinct personas, models, workspaces, and local memory, allowing for interaction either within the app or via connected messaging platforms. These bots possess the capability to read and write files, execute commands, oversee background tasks, and adhere to scheduled rules, all while managing desktop operations with adjustable permissions and approval prompts for actions beyond their designated workspace. The application accommodates multiple model providers through the Modelflare system, where billing for model usage is based on tokens instead of an AnyModelBot subscription fee. Additionally, it can be deployed in a Docker environment on a server, offering a local web console for remote access. Conversations, memory, files, and credentials are securely stored on the user's device, while requests for models are routed through Modelflare to connect with the chosen provider. With its robust features, AnyModelBot empowers users to tailor their AI experience to meet their specific needs and preferences.