# Getting Started with SOPHIAClaw

Welcome to **SOPHIAClaw** — an AI assistant built for teams who need absolute control, zero-bullshit governance, and ultra-low latency.

---

## What You're Getting

SOPHIAClaw is not another AI wrapper. It's a **multi-tiered cognitive engine** that:

- ✅ Runs on **your infrastructure** (local-first, your hardware)
- ✅ Shows **every decision** before acting (intent capture + policy checks)
- ✅ Logs **every action** immutably (complete audit trail)
- ✅ Works via **9+ channels** (Telegram, Discord, Slack, iMessage, and more)

---

## 5-Minute Quick Start

### What You Need

- **Computer:** macOS, Linux, or Windows (WSL2)
- **Node.js:** 22+
- **Time:** ~30 minutes for setup

**Hardware requirements:**

- 4GB RAM minimum (8GB recommended)
- 2GB free disk space
- Internet connection

### Installation

```bash
# Install SOPHIAClaw globally
npm install -g sophiaclaw@latest

# Run the setup wizard
sophiaclaw onboard --install-daemon
```

The wizard will:

1. Check your system requirements
2. Set up the **Gateway** (the brain of the operation)
3. Configure your first messaging channel
4. Install as a background service (optional)

### First Conversation

Once installed, start chatting:

```bash
# Start the gateway
sophiaclaw gateway --port 37521

# In another terminal, send a message
sophiaclaw agent --message "Hello, what can you do?"
```

Or connect via:

- **Telegram:** Message @YourBotName
- **Discord:** Use `/claw` command in your server
- **Web Dashboard:** http://localhost:3000

---

## Understanding What Just Happened

When you send a message to SOPHIAClaw, here's what happens behind the scenes:

1. **Intent Capture** — SOPHIA analyzes what you're asking for
2. **Policy Check** — It checks if this action is allowed (no accidental file deletions)
3. **Tool Selection** — It picks the right tool for the job
4. **Execution** — It does the work
5. **Audit Trail** — Everything is logged (you can review later)

**The key difference:** Traditional AI assistants skip step 2 and 5. SOPHIAClaw doesn't.

---

## Your First Real Tasks

### Check Your Website

```bash
sophiaclaw agent --message "Check my website getthalamus.ai and tell me if it's loading properly"
```

Behind the scenes:

- SOPHIA captures intent: "Check website status"
- Policy check: ✅ Web fetch is allowed
- Tool selected: Browser automation
- Execution: Fetches your website
- Response: "Your website is loading in 180ms with HTTP 200 OK"

### Draft an Email

```bash
sophiaclaw agent --message "Draft a professional response to this email: [paste email]"
```

### List Files

```bash
sophiaclaw agent --message "List the files in my Documents folder"
```

### Web Research

```bash
sophiaclaw agent --message "Search for best practices for AI governance in small businesses"
```

---

## Configure Your Channels

**Telegram:**

```bash
sophiaclaw channel add telegram --bot-token YOUR_BOT_TOKEN
```

**Discord:**

```bash
sophiaclaw channel add discord --bot-token YOUR_BOT_TOKEN
```

**Slack:**

```bash
sophiaclaw channel add slack --app-token YOUR_APP_TOKEN
```

See [Channel Setup](/docs/channels/) for detailed instructions.

---

## Model Configuration

By default, SOPHIAClaw uses **cost-optimized routing**:

| Task Type        | Model Used    | Cost per 1K tokens |
| ---------------- | ------------- | ------------------ |
| Simple questions | Gemini Flash  | $0.0001            |
| Most tasks       | Claude Haiku  | $0.004             |
| Complex analysis | Claude Sonnet | $0.015             |
| Code generation  | Qwen Coder    | $0.002             |

Override for specific tasks:

```bash
sophiaclaw agent --message "Complex analysis task" --model claude-3.5-sonnet
```

---

## Governance in Action

### Check Audit Trail

```bash
sophiaclaw bulletin list --last 24h
```

Shows every action SOPHIAClaw took, with:

- What was requested
- What tools were used
- What the results were
- Any policy decisions made

### Review Intent History

```bash
sophiaclaw intent list --session <session-id>
```

See how SOPHIA understood your requests over time.

### Check Policy Violations

```bash
sophiaclaw policy violations --last 7d
```

See what actions were blocked and why.

---

## Cost Control

SOPHIAClaw tracks every API call:

```bash
# Check today's spending
sophiaclaw usage today

# Check this month
sophiaclaw usage month

# Set a daily budget alert
sophiaclaw config set daily_budget_limit 10
```

Typical daily usage for a small business:

- 20-30 queries: ~$0.50-$1.00
- Heavy usage (100+ queries): ~$3.00-$5.00

---

## Next Steps

1. **Read the SMB Guide** — [FOR-SMB-OWNERS.md](/docs/FOR-SMB-OWNERS.md)
2. **Understand Governance** — [governance.md](/docs/governance.md)
3. **Learn Architecture** — [architecture.md](/docs/architecture.md)
4. **Explore Features** — [features.md](/docs/features.md)

---

## Support

- **Documentation:** https://docs.sophiaclaw.ai
- **Issues:** https://github.com/sophiaclaw/sophiaclaw/issues
- **Discord:** [Join the community](https://discord.gg/sophiaclaw)

---