basescope
以下摘录自该仓库的 README(片段) · 在 GitHub 读完整 README →
**The read-only safety layer for onchain AI agents.**
basescope is a Model Context Protocol (MCP) server that gives AI assistants — Claude, Cursor, and any MCP client — **safe, read-only** onchain abilities on **Base** and other EVM chains. Its job is to answer one question before any wallet ever signs anything:
> **"Is this token / contract / approval actually safe?"**
It checks for honeypots, rug-pull authority, hidden taxes, malicious addresses, and risky approvals — and looks up balances, ENS names, Basenames (*.base.eth), verified source, gas, and prices along the way.
Why it's different
- 🔒 **Read-only by design.** There is *no* code path that can sign or send a transaction, and **no private key can ever be configured.** It cannot move your funds because it literally has no way to. It's the inspector, not the wallet — a safe companion to transaction-capable agents.
- 🔑 **No API keys required.** Works out of the box on free, public data sources. (Optional keys can raise rate limits later, but nothing is required.)
- 🔵 **Base-first, multichain-ready.** Tuned for Base, and also supports Ethereum, Optimism, Arbitrum, and Polygon.
- 🧩 **One question, cross-checked.** Token safety is cross-referenced across **two independent sources** (GoPlus + honeypot.is) with provenance, so an agent can see when they disagree.
---
Quickstart
Add basescope to any MCP client — **Claude Desktop** (claude_desktop_config.json), **Cursor** (.cursor/mcp.json), or any other. No install step; npx fetches it on first run:
Restart your client and ask something like *"Use basescope to check if token 0x… on Base is safe."*
Prefer to run from source?
Then set the server command to node with args ["/absolute/path/to/basescope/dist/index.js"].
---
Tools
All tools are read-only. chain accepts base (default), ethereum, optimism, arbitrum, or polygon. Address fields accept a 0x address, an ENS name, **or** a Basename (*.base.eth) — Basenames resolve automatically via the Base L2 Resolver.
| Tool | What it does |
| --- | --- |
| **check_token_safety** | Honeypot / rug assessment: honeypot check, buy/sell tax, mint/blacklist/pausable/self-destruct authority, verified source, holder count → a normalized riskLevel (critical/high/medium/low). Cross-checks GoPlus + honeypot.is. |
| **check_address_safety** | Reputation check against known-malicious databases (phishing, sanctioned, scam, money-laundering, mixer, …). |
| **get_token_approvals** | Lists a wallet's outstanding ERC-20 approvals and flags risky spenders — the #1 wallet-drain vector — surfacing risky ones first. |
| **get_verified_source** | Whether a contract's source is verified (via Sourcify) + its name, compiler, and function list. |
| **inspect_contract** | Is the address a contract? Bytecode size + transaction count. |
| **get_token_info** | ERC-20 name / symbol / decimals, and optionally a holder's balance. |
| **get_native_balance** | Native coin (ETH / POL / …) balance of an address or ENS name. |