AnyModelBot

Overview

AnyModelBot is an open-source, local-first AI agent application for Windows, macOS, and Linux. Each bot has its own persona, default model, workspace, and memory, and can be reached in the app or through WeChat and Feishu. Bots can read and write files, run commands, and operate the desktop, asking for confirmation before working outside their workspace. Conversations, memories, files, and API keys stay on the device, while model requests are forwarded through Modelflare to the selected provider and billed by token usage.

Key Features

• Computer operation lets bots see the screen, open apps, read and write files, and run commands; only 24 actions are exposed, while script execution and process termination are not visible to the model, and operations outside the workspace require approval or can be switched to

Use Cases

• File and invoice organization: ask a bot to open the Downloads folder, select invoices, and move them into a reimbursement directory; operations outside the workspace require confirmation. Code review and automation: check for out-of-bounds errors, run tests or scripts, and

Getting Started

• Download and install the package for your system, or use Docker to deploy on a server. On first launch, sign in to or register for Modelflare in the browser; for Docker or remote deployment, paste an API key instead. After connecting, choose a default model, which can be changed

Deployment & Requirements

• Desktop support covers Windows, macOS, and Linux; installers cover Windows x64 and arm64, macOS Apple silicon and Intel, and Linux x64 and arm64. Docker deployment is intended for NAS or servers, with images for x64 and arm64; the web console requires an access token, and port

Before You Adopt

• License: MIT. Review its terms before using, modifying, or distributing the project.