AgentPad is a visual workflow builder for AI agents. Drag and drop nodes onto a canvas, wire them together, and run entire agent pipelines — from LLM calls to tool invocations to conditional branching — all from a single interface.
What it does
- Visual DAG editor — drag-and-drop canvas powered by SolidFlow; arrange LLM, Tool, Condition, and Output nodes, connect edges, and save/load workflows.
- Real-time execution — backend DAG executor sorts nodes topologically and streams results via SSE as each node completes.
- LLM routing —
LLMRouterdispatches to OpenAI (gpt-*,o1,o3,o4) or local Ollama (llama,mistral,qwen,deepseek, etc.) based on model name. - Three demo workflows — Research Agent, Code Reviewer, and Data Pipeline, seeded on first run.
Tech Stack
- Backend: Python, FastAPI, SQLAlchemy (aiosqlite), httpx, uvicorn
- Frontend: SolidJS, SolidFlow, Vite, Tailwind CSS
- Infra: Docker Compose, SQLite
