# Voyager AI Trip Planner - Terminal Agent with One MCP Tool URL

Voyager itself runs interactively in the terminal and uses LangChain's `ChatGroq` integration with `GROQ_API_KEY`. A quiet background gateway exposes every tool to both Voyager and MCP Inspector through the same URL:

`http://localhost:3000/mcp`

## Available tools

| Tool | Data source |
|---|---|
| `search_destinations` | Local JSON database |
| `get_weather` | Open-Meteo API |
| `find_attractions` | Wikipedia API |
| `estimate_trip_budget` | Python calculation |

The same server also exposes the `destination://{name}` resource and the `plan_a_trip` prompt.

## Setup

```powershell
cd C:\Users\ASUS\mcp\trip-planner
python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
Copy-Item .env.example .env
```

Add your Groq key to `.env`:

```env
GROQ_API_KEY=gsk_your_real_key_here
GROQ_MODEL=llama-3.3-70b-versatile
```

## Run Voyager in the terminal

The launcher silently starts the tool gateway and then opens the interactive Voyager agent in the current terminal:

```powershell
.\.venv\Scripts\python.exe run_project.py
```

Only the agent UI is shown. When Voyager exits, the hidden tool gateway stops automatically.

## View all tools in one MCP Inspector connection

If Voyager is already running through `run_project.py`, leave it running and connect Inspector directly. Otherwise, start only the tool gateway for Inspector testing:

```powershell
.\.venv\Scripts\python.exe server.py
```

In another terminal, start Inspector:

```powershell
npx @modelcontextprotocol/inspector
```

Choose **Streamable HTTP**, enter `http://localhost:3000/mcp`, and connect. All four tools appear under that single connection. The agent itself is not served at this URL; it remains a terminal program.

## Run only the agent

If `server.py` is already running:

```powershell
.\.venv\Scripts\python.exe agent.py
```

Weather and attraction tools require internet access. Budget values are demonstration estimates, not live booking quotes.