Designing Real-World AI Agents Workshop

Clone the workshop repository to your local machine. The repo contains all agent code, configuration templates, slides, and Make targets.

git clone https://github.com/iusztinpaul/designing-real-world-ai-agents-workshop.git
cd designing-real-world-ai-agents-workshop

Copy the example .env file and fill in your API keys. GOOGLE_API_KEY is mandatory; OPIK_API_KEY is optional but recommended for tracking agent runs.

cp .env.example .env
# Edit .env and set:
# GOOGLE_API_KEY=your_google_ai_studio_key
# OPIK_API_KEY=your_opik_key  (optional)

The project uses uv for fast, reproducible Python dependency management. Run uv sync to install all packages defined in pyproject.toml.

uv sync

The Makefile provides a test target that exercises both agents end-to-end. Passing tests confirm your API keys and dependencies are correctly configured.

make test-end-to-end

Each agent is served as a separate MCP server. Add both the Deep Research server and the LinkedIn Writing server to your claude_desktop_config.json pointing to their respective module entry points.

{
  "mcpServers": {
    "deep-research-agent": {
      "command": "uv",
      "args": ["run", "python", "-m", "src.deep_research.mcp_server"],
      "cwd": "/path/to/designing-real-world-ai-agents-workshop",
      "env": {
        "GOOGLE_API_KEY": "your_google_api_key"
      }
    }
  }
}

The workshop supports four ways to run the agents: as MCP servers (for Claude Desktop integration), as slash-command skills, through a Streamlit web UI, or directly via Make targets. Start with the Streamlit UI for an interactive introduction before wiring up the MCP servers.

# Launch Streamlit UI
make streamlit