# Install RobotBase MCP (three editions)

RobotBase exposes three remote MCP servers. Pick the one that matches the agent; all use
Streamable HTTP and need no API key.

| Edition | URL | Use when |
|---|---|---|
| Web3 | `https://web3.robotbase.cc/mcp` | on-chain verification, risk, prediction/crypto trends |
| Dev | `https://dev.robotbase.cc/mcp` | coding: GitHub/HF/npm/PyPI/HN, web docs tooling |
| Content | `https://content.robotbase.cc/mcp` | content/SEO: web, social hot lists, attention trends, books |

## Claude Desktop / Cursor / Cline / VS Code / Codex

```json
{
  "mcpServers": {
    "robotbase-web3":   { "type": "http", "url": "https://web3.robotbase.cc/mcp" },
    "robotbase-dev":    { "type": "http", "url": "https://dev.robotbase.cc/mcp" },
    "robotbase-content":{ "type": "http", "url": "https://content.robotbase.cc/mcp" }
  }
}
```

Claude Code:

```bash
claude mcp add --transport http robotbase-web3 https://web3.robotbase.cc/mcp
claude mcp add --transport http robotbase-dev https://dev.robotbase.cc/mcp
claude mcp add --transport http robotbase-content https://content.robotbase.cc/mcp
```

## Verify

```bash
curl -s https://web3.robotbase.cc/mcp/healthz
# {"ok": true, "tools": 65, "version": "0.32.3"}
curl -s https://dev.robotbase.cc/mcp/healthz
# {"ok": true, "tools": 18, "version": "0.32.3"}
curl -s https://content.robotbase.cc/mcp/healthz
# {"ok": true, "tools": 35, "version": "0.32.3"}
```

## Free tier and payment

- Web tools: 100 calls/day/IP per edition, then 1–3 x402 credits.
- Web3 risk: 5 calls/day/IP; Content CyberLibrary: 5 calls/day/IP.
- Credits: 0.01 USDC each; 10 / 0.10 USDC, 100 / 1.00 USDC; USDC on Base or Solana.
- No API key, no account, no subscription. Retry a 402 with the same x402 payment payload;
  credit balance is shared across editions.

## Troubleshooting

- Refresh the client's tool list after switching editions; each endpoint has its own catalog.
- `robotbase.cc/mcp` is a Web3 compatibility alias; use `web3.robotbase.cc` for the canonical URL.
- Public pages only; `web_fetch` respects robots.txt and does not bypass logins/paywalls.