Live AI Agents
Thinking in Real-Time.

An open-source crypto experiment tracking live token deployments, funding rates, arbitrage gaps, and prediction markets—streaming dynamic analysis generated by Claude 3 directly to your browser 24/7.

Enter Live Dashboard Fork The Code

Under the Hood

This project bridges live data sources with Claude 3 Haiku, streaming the generated agent actions natively to a frontend dashboard via WebSockets.

Live Data Integration

Agents scrape real-time context. The Arb Agent hits Binance/Uniswap APIs, the Futures Agent grabs live funding rates, and the Meme Hunter scans Solana Raydium Pools.

Claude 3 Synthesis

Live data is instantly piped into Anthropic's endpoints. Claude dynamically synthesizes distinct, 1-sentence thought processes and action plans for each specific agent loop.

WebSocket Streaming

Instead of reloading pages, FastAPI orchestrates persistent `wss://` connections that natively type out the agent's logic into the dashboard terminal 24/7.

Core Design Stack

FastAPI Backend (Railway)

Written in modern Python. Employs `asyncio` to simultaneously maintain dozens of non-blocking WebSockets without breaking a sweat while securely obfuscating the Anthropic API keys.
Anthropic Python SDK

Connected natively to `claude-3-haiku-20240307` for blisteringly fast latency. Built with robust fallback generation if rate limiting occurs.
Vercel Edge Frontend

The static HTML, Tailwind CSS styling, and client-side JavaScript engine are cached globally. Clean URLs routing ensures smooth navigation between the docs and the dashboard.

# Project Structure
experiments.html             # Frontend layout for WebSocket consumption
index.html                   # Landing + Marketing Copy
vercel.json                  # Clean URL Vercel overrides
backend_api/                 # The Server
  backend.py              # FastAPI + Asyncio Event loops
  requirements.txt        # Python dependencies
  Procfile                # Railway deploy targets

Quick Start Local Dev

                    Terminal
                    
                
# 1. Clone
$ git clone https://github.com/888BasedGod-sol/autoagent.git
# 2. Setup Backend
$ cd autoagent/backend_api
$ pip install -r requirements.txt
# 3. Add Claude Key (Optional, will fallback if missing)
$ echo 'ANTHROPIC_API_KEY="sk-..."' > .env
# 4. Boot Python Server
$ uvicorn backend:app --reload --port 8000
# 5. Boot Frontend Host
$ python3 -m http.server 8080

Dive In Now

Stop reading the docs and go watch the agents stream live.

> /experiments

Live AI Agents Thinking in Real-Time.