AnyModelBot
Local-first AI agent app that connects any model to your desktop, WeChat, and Feishu.
Open-source alternative to Grok Bot and Claude Duet running on your machine via Modelflare.
AnyModelBot is a local-first AI agent application for Windows, macOS, and Linux that runs on your computer or server. It lets you create custom bots with specific personas, local workspaces, and memories while using 70+ models through a single Modelflare account. Professionals can interact with their bots directly via the desktop app, WeChat, Feishu, Telegram, Slack, or Discord.
What it is
The application functions as a persistent personal AI agent that can read and write files, execute commands in a sandbox, operate your desktop via Cua Driver, and run background tasks or scheduled rules. It features built-in permission tiers so that any action outside a bot's designated workspace pauses to request your approval. All conversations, file data, and memory files remain stored locally on your device in a SQLite database and Markdown files.
Who it's for
It suits professionals, developers, and power users who want an autonomous local AI assistant capable of interacting with local files and chat apps without relying on proprietary cloud subscriptions. It does not suit users looking for a fully managed, zero-setup cloud service or those uncomfortable running desktop binaries and configuring API routing.
Key features
Multi-Model Support
Access GPT, Claude, Gemini, Grok, DeepSeek, and 70+ other models through a single Modelflare account.
Local Workspace Control
Bots operate within isolated folders with three permission tiers and require your approval for out-of-bounds actions.
Chat Channel Integration
Connect your bots to WeChat, Feishu, Telegram, Slack, and Discord alongside the local desktop interface.
Background Tasks & Rules
Set up timed rules and long-running background tasks that notify you only when they require a decision.
Up and running in 3 steps
Our honest verdict
What's great
- Open-source and local-first architecture keeps personal data, chats, and workspace files on your device.
- Supports multiple communication channels including WeChat, Feishu, Telegram, Slack, and Discord.
- Built-in permission system pauses out-of-bounds file or desktop actions until you approve them.
- Provides granular budget caps for individual automated background tasks and recurring rules.
Where it falls short
- Requires an external Modelflare account and API key usage balance to run model inference.
- Desktop computer operation features are currently in Beta on macOS.
- Requires Node.js 22.13+ and Go 1.26.8+ if you choose to build and run the application from source.
Pricing & license
The software is free under its MIT license. Model usage is paid per token through a Modelflare account, with no secondary application subscriptions required.
Open remote AI-training roles
879 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Where are my conversation histories and files stored?
All conversations, local SQLite databases, workspace files, and Markdown memory files stay entirely on your device.
How does the bot handle actions outside its workspace?
When a bot attempts to read, write, or execute commands outside its assigned folder, it halts execution and presents an approval card for you to accept or deny.
Do I need a paid subscription to use AnyModelBot?
The application itself is free and open source, but you pay per use for model tokens through your Modelflare account.
