beefree-agent-demo¶
A PydanticAI agent that drives the Beefree SDK via MCP, with a built-in mock server so you can run the full loop locally with no Beefree account and no beta access required.
Model-agnostic by design. Most agent demos are secretly OpenAI demos with a provider abstraction bolted on at the end. This one starts from pydantic-ai's provider-neutral Agent, and switching between Anthropic, OpenAI, and Gemini is one env-var change. There is no OpenAI-shaped API call anywhere in the codebase.
What it does¶
┌─────────────┐ WebSocket ┌──────────────────┐ MCP/JSON-RPC ┌──────────────────┐
│ Vite + React│ ←──────────→ │ FastAPI + Pydantic│ ←─────────────→ │ Beefree MCP │
│ (port 5173) │ │ AI Agent │ │ (mock: 8001, │
│ │ │ (port 8000) │ │ real: your URL)│
└─────────────┘ └──────────────────┘ └──────────────────┘
- Left panel - chat with the agent
- Right panel - live JSON preview of the Beefree template the agent built or fetched
- Agent tools -
create_email,update_section,list_templates,get_template
Stack¶
| Layer | Tech |
|---|---|
| Frontend | Vite, React, TypeScript, TanStack Router, HeroUI, Tailwind v4, Zustand, Zod |
| Backend | FastAPI, PydanticAI, Python 3.13+ |
| Package managers | Bun (frontend), uv (backend) |
| MCP transport | Streamable HTTP (JSON-RPC 2.0) |
Get started¶
See the Quickstart for setup instructions, or jump to Configuration to understand all the knobs.