# arnfa-mcp

MCP server for **[Arnfa](https://arnfa.vercel.app)** — weather-aware trip planning across Thailand. Lets any AI agent read the Thai sky and plan a weather-fit day in 150+ areas (Bangkok neighbourhoods & districts, all 76 province cities, tourist hotspots).

Reads Arnfa's open API — **no key, no fabrication** (real forecast or an honest error).

## Tools
- `arnfa_list_areas(query?, tier?)` — find a valid area key (search by Thai/English name).
- `arnfa_plan_day(area, budget?, day?)` — an ordered, weather-fit itinerary; `day` 0–6 plans ahead.
- `arnfa_sky_now(area)` — current sky + a go-out verdict for one area.
- `arnfa_where(day?)` — rank WHERE the sky is clearest across Thailand ("ไปไหนดี").
- `arnfa_cooling_centers(lat, lng, n?)` — nearest official Bangkok cooling refuges (ห้องหลบร้อน) for dangerous heat / PM2.5.
- `arnfa_ask(message, prior?)` — ask in natural Thai/English; the grounded engine reads the real sky and narrates a plan. Pass back the previous reply's `intent` as `prior` to refine (multi-turn).

## Quickstart
```bash
npx arnfa-mcp        # runs the stdio MCP server, no install
```

Wire into a client (e.g. Claude Code `.mcp.json`):
```json
{ "mcpServers": { "arnfa": { "command": "npx", "args": ["-y", "arnfa-mcp"] } } }
```

## From source
```bash
npm install && npm run build
node dist/index.js                                # stdio MCP server
ARNFA_BASE=http://localhost:3248 npm run smoke    # quick check against a deployment
```

`ARNFA_BASE` (default `https://arnfa.vercel.app`) points the tools at any Arnfa deployment.

MIT.