Lesson 5 — Extending Your Agent

Your research agent is working. Here's how to make it 10x more powerful:

Your research agent is working. Here's how to make it 10x more powerful:

Add a write_file tool:

def write_file(filename: str, content: str) -> str:

Now your agent can save reports to disk automatically. Add it to your TOOLS list and your agent will start filing its own research.

**Add a summarize_and_score tool:** Have the agent score each source on relevance (1–10) before deciding whether to read the full page. This makes it faster and more selective.

**Add memory:** Store past research sessions in a JSON file. On startup, load previous context so the agent can build on past work instead of starting from scratch.

**Build a web UI:** Ask Claude to wrap your agent in a FastAPI backend with a simple streaming frontend. Users type a question, watch the agent work in real time, and get a formatted report. This is a deployable product.

**What this is worth:** Research automation agents for industries like real estate, recruiting, finance, and e-commerce sell for $2,000–$20,000 to build. Monthly subscriptions for ongoing research services run $200–$2,000/month. You built the core in under an hour.

The next step is learning multi-agent systems — where multiple specialized agents collaborate, hand off work, and check each other's outputs. That's where the real leverage is. See you in the next course.