# Examples

Real-world examples of using llm.energy.

!!! tip "Try It Live"
    All examples can be tested at [llm.energy](https://llm-energy-lp642k3kpa-uc.a.run.app) - just enter the URL!

---

## Extracting Anthropic Docs

```bash
# Web App
https://llm-energy-lp642k3kpa-uc.a.run.app

# Enter URL
docs.anthropic.com
```

**Result:** 15+ markdown files covering Claude API, prompt engineering, and more.

---

## Extracting MCP Docs

```bash
# URL
modelcontextprotocol.io
```

**Result:** Complete MCP specification, server implementation guides, and client examples.

---

## Extracting Stripe Docs

```bash
# URL
docs.stripe.com
```

**Result:** API reference, integration guides, and webhook documentation.

---

## Using with Claude

1. Extract documentation using llm.energy
2. Download the ZIP archive
3. In Claude, start a new conversation
4. Upload `llms-full.md` or individual files
5. Ask questions about the documentation

**Example prompt:**

> "Based on the uploaded documentation, how do I implement streaming responses?"

---

## Using with ChatGPT

1. Extract documentation
2. Download files
3. Upload to ChatGPT (GPT-4 with file upload)
4. Reference the docs in your prompts

---

## Using install.md

### Extracting install.md

When extracting from a site that has install.md, llm.energy will automatically detect and include it:

```bash
# URL with install.md
python.langchain.com
```

**Result:** Documentation files + `install.md` with installation instructions.

### Creating install.md

Use the [install.md Generator](/install-generator) to create your own:

1. Go to the Generator page
2. Fill in product name, objective, and done criteria
3. Add TODO items (checklist for the LLM)
4. Add installation steps with code blocks
5. Download the generated `install.md`

### Using install.md with AI Agents

Give an AI agent the install.md file and ask it to install the software:

**Example prompt:**

> "Follow the instructions in install.md to set up this tool on my machine"

The agent will:

1. Read the OBJECTIVE to understand the goal
2. Execute each step in the TODO list
3. Verify completion against DONE WHEN criteria

---

## Using the MCP Server

With the MCP server, AI assistants can extract docs programmatically:

**Claude Desktop**

```
You: "Extract the Vercel documentation and tell me about edge functions"

Claude: [Uses extract_documentation tool]
        [Reads relevant sections]
        "Based on the Vercel docs, edge functions are..."
```

**Example conversation:**

```
You: "What MCP tools do you have for documentation?"

Claude: "I have access to llm-energy with these tools:
        - extract_documentation: Extract docs from a URL
        - fetch_llms_txt: Get raw llms.txt content
        - get_document: Retrieve a cached document
        - list_documents: List all cached docs"

You: "Extract docs from docs.anthropic.com"

Claude: [Extracts and summarizes the documentation]
```

---

## Supported Sites

Sites known to have llms.txt support:

| Site | URL |
|------|-----|
| Anthropic | docs.anthropic.com |
| MCP | modelcontextprotocol.io |
| Stripe | docs.stripe.com |
| Vercel | vercel.com/docs |
| Supabase | supabase.com/docs |
| Mintlify | mintlify.com/docs |

!!! tip "Check any site"
    Try adding `/llms.txt` or `/llms-full.txt` to any documentation URL to see if it's supported.