Search, SERP, scrape, extract, map, crawl, verify, discover, research — plus 6 proxy types — through a single Streamable-HTTP MCP endpoint. Drop into Claude Desktop, Claude Code, Cursor, Codex, Cline, or Windsurf in 30 seconds. 搜索、SERP、抓取、提取、站点地图、爬取、验证、发现、研究,加上 6 种代理 — 全部通过一个 Streamable-HTTP MCP 端点。30 秒接入 Claude Desktop、Claude Code、Cursor、Codex、Cline、Windsurf。
https://mcp.novada.com/YOUR_API_KEY/mcp
Works in Claude.ai, Claude Desktop, Cursor, and all MCP clients. Replace YOUR_API_KEY with your Novada API key.
适用于 Claude.ai、Claude Desktop、Cursor 等所有 MCP 客户端。将 YOUR_API_KEY 替换为你的 Novada API 密钥。
1,000 tool calls / month free · No credit card · 60 req/min per IP 每月 1,000 次免费调用 · 无需信用卡 · 每 IP 60 次/分钟
No install needed. Verify the endpoint, then wire up your client. 无需安装。先验证端点能正常响应,再接入客户端。
$ curl -sX POST https://mcp.novada.com/mcp \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
Expected response: a JSON-RPC envelope with result.tools listing 25 tool descriptors. If you see a 501 for novada_browser*, that's expected — those are local-only.
期望响应:JSON-RPC 包装,result.tools 列出 25 个工具描述符。novada_browser* 返回 501 属于预期(仅本地可用)。
Pick your client. Copy the snippet. Paste it. Replace YOUR_API_KEY with your real key.
选择客户端,复制片段,粘贴。把 YOUR_API_KEY 替换为你的真实密钥。
Option A — Custom Connector UI (Free, Pro, Max). In Claude Desktop go to Customize → Connectors → + → Add custom connector and paste the URL. Free tier is limited to one custom connector. 方式 A — 自定义连接器界面(Free、Pro、Max 均可)。在 Claude Desktop 中:Customize → Connectors → + → Add custom connector,粘贴下方 URL。Free 版限一个自定义连接器。
Team / Enterprise owners: Organization settings → Connectors → Add → Custom → Web first, then members can connect. Team / Enterprise 管理员:先在 Organization settings → Connectors → Add → Custom → Web 添加,成员才能连接。
Option B — stdio bridge. Edit claude_desktop_config.json and bridge via mcp-remote:
方式 B — stdio 桥接。编辑 claude_desktop_config.json,使用 mcp-remote 桥接:
{
"mcpServers": {
"novada": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.novada.com/mcp?token=YOUR_API_KEY"
]
}
}
}
Config file: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) · %APPDATA%\Claude\claude_desktop_config.json (Windows). The -y auto-accepts the npx install prompt.
配置文件:~/Library/Application Support/Claude/claude_desktop_config.json (macOS) · %APPDATA%\Claude\claude_desktop_config.json (Windows)。-y 用于自动确认 npx 安装。
Recommended — pass the key as a header, scope it to your user so every project sees it: 推荐方式 — 使用请求头传递密钥,并设置为用户级作用域,所有项目可见:
$ claude mcp add --transport http novada \ https://mcp.novada.com/mcp \ --header "Authorization: Bearer YOUR_API_KEY" \ --scope user
Or the one-liner URL form (token visible in shell history): 或使用单行 URL 形式(密钥会进入 shell 历史):
$ claude mcp add --transport http novada \ https://mcp.novada.com/mcp?token=YOUR_API_KEY --scope user
Verify with claude mcp list or claude mcp get novada. Without --scope user the server is registered only in the current project directory.
用 claude mcp list 或 claude mcp get novada 验证。不加 --scope user 时,服务器仅在当前项目目录内可用。
Add to ~/.cursor/mcp.json:
添加到 ~/.cursor/mcp.json:
{
"mcpServers": {
"novada": {
"url": "https://mcp.novada.com/mcp?token=YOUR_API_KEY"
}
}
}
Project-scoped alternative: drop the same JSON at .cursor/mcp.json in the repo root. After saving, open Cursor Settings (⌘⇧J) → Features → Model Context Protocol and confirm Novada toggles green.
也可放在仓库根目录的 .cursor/mcp.json(项目级)。保存后打开 Cursor Settings(⌘⇧J)→ Features → Model Context Protocol,确认 Novada 显示为绿色。
Add to ~/.codex/config.toml:
添加到 ~/.codex/config.toml:
[mcp_servers.novada] url = "https://mcp.novada.com/mcp?token=YOUR_API_KEY"
Codex auto-detects Streamable HTTP from the presence of url (stdio servers use command+args instead). To keep the key out of the file, use bearer_token_env_var = "NOVADA_API_KEY" against https://mcp.novada.com/mcp instead. Source: developers.openai.com/codex/config-reference.
Codex 通过 url 字段自动识别 Streamable HTTP(stdio 服务器使用 command+args)。若不希望将密钥写入文件,可改用 bearer_token_env_var = "NOVADA_API_KEY" 搭配 https://mcp.novada.com/mcp。来源:developers.openai.com/codex/config-reference。
Edit cline_mcp_settings.json (Cline → MCP Servers → Edit):
编辑 cline_mcp_settings.json(Cline → MCP Servers → 编辑):
{
"mcpServers": {
"novada": {
"type": "streamableHttp",
"url": "https://mcp.novada.com/mcp?token=YOUR_API_KEY"
}
}
}
"type": "streamableHttp" must be explicit — without it, Cline defaults remote servers to SSE for backwards compatibility. The Cline SDK/CLI also accepts a nested "transport": { "type": "streamableHttp", "url": "..." } form. If you have a legacy config with "transportType": "http", current Cline auto-upgrades it.
"type": "streamableHttp" 必须显式写出 — 否则 Cline 为了向后兼容会默认走 SSE。Cline SDK / CLI 也接受嵌套形式 "transport": { "type": "streamableHttp", "url": "..." }。旧配置中若使用 "transportType": "http",当前 Cline 会自动升级。
Edit ~/.codeium/windsurf/mcp_config.json:
编辑 ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"novada": {
"serverUrl": "https://mcp.novada.com/mcp?token=YOUR_API_KEY"
}
}
}
Windsurf's documented field is serverUrl (url also works as a fallback). After saving, press the refresh button in the MCP panel to load the new server.
Windsurf 文档示例使用 serverUrl(也兼容 url)。保存后,在 MCP 面板按 刷新按钮 加载新服务器。
Using ChatGPT? Custom remote MCP requires Developer mode in Settings → Connectors and may need OAuth on some plans — check current ChatGPT docs. 使用 ChatGPT?远程 MCP 需在 设置 → 连接器 中开启 开发者模式,部分套餐可能要求 OAuth — 查看 ChatGPT 最新文档。
Building agents in code? Point any MCP-aware framework at the same endpoint. Configs verified against each project's current docs. 用代码构建 Agent?让任意支持 MCP 的框架指向同一端点。以下配置均对照各项目最新文档核对。
Python — pip install langchain-mcp-adapters. Native Streamable HTTP.
Python — pip install langchain-mcp-adapters。原生支持 Streamable HTTP。
from langchain_mcp_adapters.client import MultiServerMCPClient client = MultiServerMCPClient({ "novada": { "transport": "http", "url": "https://mcp.novada.com/mcp", "headers": {"Authorization": "Bearer YOUR_NOVADA_API_KEY"}, } }) tools = await client.get_tools()
Python — pip install llama-index-tools-mcp. Auto-detects Streamable HTTP.
Python — pip install llama-index-tools-mcp。自动识别 Streamable HTTP。
from llama_index.tools.mcp import BasicMCPClient, McpToolSpec client = BasicMCPClient( "https://mcp.novada.com/mcp", headers={"Authorization": "Bearer YOUR_NOVADA_API_KEY"}, ) tools = await McpToolSpec(client=client).to_tool_list_async()
Python — pip install openai-agents. Native MCPServerStreamableHttp.
Python — pip install openai-agents。原生 MCPServerStreamableHttp。
from agents.mcp import MCPServerStreamableHttp server = MCPServerStreamableHttp( name="Novada MCP", params={ "url": "https://mcp.novada.com/mcp", "headers": {"Authorization": "Bearer YOUR_NOVADA_API_KEY"}, }, cache_tools_list=True, ) async with server: ... # pass to Agent(mcp_servers=[server])
Python — pip install crewai-tools. Via MCPServerAdapter.
Python — pip install crewai-tools。通过 MCPServerAdapter。
from crewai_tools import MCPServerAdapter params = { "url": "https://mcp.novada.com/mcp", "transport": "streamable-http", "headers": {"Authorization": "Bearer YOUR_NOVADA_API_KEY"}, } with MCPServerAdapter(params) as tools: agent = Agent(..., tools=tools)
If the streamable-http passthrough is flaky on your version, bridge with npx mcp-remote https://mcp.novada.com/mcp --header "Authorization: Bearer YOUR_NOVADA_API_KEY" and point CrewAI at the local SSE URL.
若你的版本对 streamable-http 透传不稳定,可用 npx mcp-remote https://mcp.novada.com/mcp --header "Authorization: Bearer YOUR_NOVADA_API_KEY" 桥接,再让 CrewAI 指向本地 SSE 地址。
MCP Client Tool node. n8n's node speaks SSE — bridge the Streamable-HTTP endpoint with mcp-remote:
MCP Client Tool 节点。n8n 节点走 SSE — 用 mcp-remote 桥接 Streamable-HTTP 端点:
$ npx mcp-remote https://mcp.novada.com/mcp \ --header "Authorization: Bearer YOUR_NOVADA_API_KEY"
Then in the node set SSE Endpoint → the local bridge URL (e.g. http://localhost:3000/sse). Or, if your n8n build supports it, paste https://mcp.novada.com/mcp?token=YOUR_NOVADA_API_KEY directly with Bearer auth credential.
然后在节点的 SSE Endpoint 填本地桥接地址(如 http://localhost:3000/sse)。若你的 n8n 版本支持,也可直接填 https://mcp.novada.com/mcp?token=YOUR_NOVADA_API_KEY 并配 Bearer 凭据。
Just paste a URL. No npm, no env vars, no rebuilds. Works on any machine in 30 seconds. 只需粘贴一个 URL,无需 npm、无需环境变量、无需重新构建。任何机器 30 秒接入。
Tools update server-side. Your agents always run the newest version — new features and fixes propagate without a redeploy on your side. 工具在服务端更新,你的 Agent 始终运行最新版本 — 新功能和修复无需你重新部署即可生效。
Runs on Vercel Edge with global PoPs. The endpoint is the same for every client; latency tracks the user, not the data center. 运行于 Vercel Edge,具备全球节点。所有客户端共用同一端点,延迟随用户位置就近响应。