Agentspan Demo Suite

A collection of working demos showing what you can build with Agentspan — durable AI agents that survive process restarts, multi-agent pipelines, and human-in-the-loop workflows.

Demos

Demo	What it shows
01 Weather Agent	Durable execution — agent survives process restarts
02 Research Pipeline	Multi-agent orchestration — 4 agents in sequence
03 Fraud Review	Human-in-the-loop — AI flags, human decides

Prerequisites

• Python 3.10+
• Agentspan server running locally at http://localhost:6767
• OpenAI API key configured in Agentspan

Setup

1. Clone the repo

git clone https://github.com/YOUR_USERNAME/agentspan-demo
cd agentspan-demo

2. Create a virtual environment

python -m venv agentspan-env
source agentspan-env/bin/activate

3. Install dependencies

pip install agentspan requests

4. Start the Agentspan server

Make sure your local Agentspan server is running at http://localhost:6767 with an OpenAI API key configured:

agentspan server start
agentspan credentials set OPENAI_API_KEY sk-...

See agentspan.ai/docs for full setup instructions.

Running the demos

Demo 1 — Weather Agent

cd demos/01_weather_agent
python weather_agent.py

Demo 2 — Research Pipeline

cd demos/02_research_pipeline
python research_pipeline.py "AI agents in production 2026"

Demo 3 — Fraud Review (Human-in-the-Loop)

cd demos/03_fraud_review
python fraud_review.py

See each demo's README.md for full details and example scenarios.

Notes

• All demos require the Agentspan server running locally
• Demo 2 saves output to demos/02_research_pipeline/output/
• Demo 3 uses interactive terminal input for approve/deny