# API Key Setup Guide

Every lab needs at least one LLM API key. If you don't have one, the mock LLM handles everything offline.

---

## Quick Start: Use the Mock LLM (No Key Needed)

All labs include automatic mock LLM fallback. Run without any setup:

```bash
python labs/L1-first-agent/solution.py data.txt "Summarize this file"
```

The mock client returns realistic responses. Your code runs identically with real or mock LLM.

---

## Provider Keys

### Anthropic (Claude) — Recommended

**Best for**: All course labs. Claude Sonnet is the default model used throughout.  
**Cost**: Free credits available on signup. Labs cost ~$0.01-0.05 each with Sonnet.  
**Get key**: [console.anthropic.com](https://console.anthropic.com/settings/keys)  
**Set it**:
```bash
# Windows PowerShell
$env:ANTHROPIC_API_KEY="sk-ant-..."

# Mac/Linux
export ANTHROPIC_API_KEY="sk-ant-..."
```

### OpenAI (GPT)

**Best for**: Labs using the OpenAI SDK.  
**Cost**: Free credits on signup. GPT-4o costs ~$0.02-0.10 per lab.  
**Get key**: [platform.openai.com/api-keys](https://platform.openai.com/api-keys)  
**Set it**:
```bash
export OPENAI_API_KEY="sk-..."
```

### Google Gemini

**Best for**: Cost-sensitive experimentation. Gemini Flash is $0.15/M tokens.  
**Cost**: Free tier available (60 requests/minute).  
**Get key**: [aistudio.google.com/apikey](https://aistudio.google.com/apikey)  
**Set it**:
```bash
export GEMINI_API_KEY="..."
```

### OpenRouter (Multi-Provider)

**Best for**: Accessing multiple models from one key.  
**Cost**: Pay-as-you-go. Free tier available for some models.  
**Get key**: [openrouter.ai/keys](https://openrouter.ai/keys)  
**Set it**:
```bash
export OPENROUTER_API_KEY="sk-or-..."
```

---

## Which Key Should You Get?

| If you... | Get |
|-----------|-----|
| Want the simplest setup | **None** (use mock LLM) |
| Want to follow the course exactly | **Anthropic** (Claude Sonnet) |
| Already have OpenAI access | **OpenAI** (GPT-4o or o3-mini) |
| Want the cheapest option | **Gemini** (Flash is $0.15/M) |
| Want to try different models | **OpenRouter** (one key for everything) |

---

## Testing Your Key

```bash
# Anthropic
python -c "from anthropic import Anthropic; c=Anthropic(); print(c.messages.create(model='claude-sonnet-4', max_tokens=50, messages=[{'role':'user','content':'hi'}]).content)"

# OpenAI
python -c "from openai import OpenAI; c=OpenAI(); print(c.chat.completions.create(model='gpt-4o', messages=[{'role':'user','content':'hi'}]).choices[0].message.content)"

# Mock (no key needed)
python -c "from mock_llm import MockAnthropic; c=MockAnthropic(); print(c.messages.create(messages=[{'role':'user','content':'hi'}]).content[0].text)"
```

---

## Free Tier Limits

| Provider | Free Credits | Rate Limit |
|----------|-------------|------------|
| Anthropic | $5-10 signup credit | Varies by model |
| OpenAI | $5-18 signup credit | Varies by tier |
| Gemini | Free tier (60 req/min) | 60 requests/minute |
| OpenRouter | Varies by model | Varies |

The course uses ~$0.50-2.00 in API costs total with Anthropic, or $0 with the mock LLM.