# token-risk-check — MCP server

A Model Context Protocol (MCP) server that exposes one sellable tool:
**`analyze_token_risk`** — a live scam/rug-pull check for any memecoin token.

AI agents call it before buying or promoting a token, and get back:

- a verdict (`PASS` / `REJECT` / `UNAVAILABLE`)
- reasons & warnings (honeypot pattern, age, buy pressure, missing socials…)
- a **RugCheck** risk report for Solana (score, top-holder %, mint/freeze authority)
- live market data from DexScreener (price, liquidity, volume, buys/sells…)

All data sources are free and keyless (DexScreener API + RugCheck API).

## Tool

| Tool | Description |
|------|-------------|
| `analyze_token_risk(contract_address, chain="solana", min_buys_5m=50)` | Assess whether a token is likely a scam or rug pull. |

## Run locally

```bash
pip install -r requirements.txt
python mcp_server.py
```

Endpoint: `http://localhost:8000/mcp` (streamable HTTP, stateless — no session
handshake needed). Test it:

```bash
curl -X POST http://localhost:8000/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'
```

## Deploy for free on Smithery (always-on, no server to run)

1. Push this folder to a GitHub repo.
2. Sign up at **https://smithery.ai**, connect GitHub, add the repo.
3. Smithery builds the `Dockerfile` (see `smithery.yaml`), runs the server,
   and gives you a permanent public MCP URL — free.
4. Use that URL in any MCP client, or submit it to marketplaces
   (MCP-Hive, Glama, mcp.so) to sell the tool per invocation.

The `PORT` env var is set by Smithery at runtime; `mcp_server.py` reads it
(8000 locally).

> This folder is self-contained — it includes its own copies of the bot's
> data + scam modules (`config.py`, `dexscreener.py`, `scam_filter.py`).
> The token-risk check is for entertainment/education only; new coins are
> overwhelmingly scams. Not financial advice.