Build AI Agents Visually
Drag, drop, and connect AI models, integrations, and Python scripts on a visual canvas. No frameworks, no complexity — just powerful agents in minutes.
OpenAI, Anthropic, Gemini, Groq, DeepSeek, and more
Email, Telegram, Discord, Slack, WhatsApp, and more
Write scripts to extend your agents with any logic
Download as a standalone Python project
Create your first agent to get started!
AI agents published by the community
Build AI agents in minutes.
Click "Create New Agent" for an AI pipeline, or "Create a Model" to fine-tune, merge, or run a local AI model.
Click "Templates" for a ready-made agent, or drag AI models, integrations, and tools from the left sidebar.
Orange dot = output. Blue dot = input. Drag from orange to blue. Drop anywhere near the dot — it snaps!
Click "Keys" for global keys (OpenAI, Anthropic, etc). Or click "Variables" on any node for per-node config.
In the sidebar, click "+ New Dataset" to create training data, or "Upload File" to import from your computer (JSON, CSV, JSONL). Your datasets appear in the Fine-Tune node dropdown.
Drag a "Fine-Tune Model" node onto the canvas. Pick a base model, select your dataset from the dropdown, choose LoRA/QLoRA/Full training, then export and run on your computer — or train right in the browser with WebGPU.
Use "Create a Model" → "Run Locally (WebGPU)" to run Llama, Phi, or Gemma right in your browser. No API key needed — everything runs on your GPU. First load downloads the model (~500MB-2GB), then it's cached.
Use "Chat" to talk to your agent. "AI Helper" builds agents automatically. "Console" shows logs. Click "Export" to download a standalone Python project.
Sign in with your email to save agents to the cloud. Works offline too — agents always save to your browser.