Skip to main content
Give AI agents real-time web data by connecting to Nimble through the Smithery MCP registry — no manual server configuration needed. Smithery is an MCP registry that hosts and manages MCP server connections. The nimble/nimble-mcp package exposes all Nimble tools — search, extract, crawl, map, and agents — to any MCP-compatible client.

​
Prerequisites

npm install -g @smithery/cli@latest
Set environment variables: 
export NIMBLE_API_KEY="your-nimble-api-key"
Get a Nimble API key from the dashboard (free trial available).

​
Quick Start — CLI

Connect to the Nimble MCP server using the Smithery CLI in three steps.
1

Create a namespace

smithery namespace create {your-namespace}
2

Add the Nimble MCP server

smithery mcp add nimble/nimble-mcp --headers '{"NIMBLE_API_KEY":"Bearer your-NIMBLE_API_KEY"}'
3

Call tools

List available tools and call them directly from the CLI:
smithery tool list {connection}
smithery tool call {connection} {tool_name} '{"key": "value"}'

​
Programmatic Usage

Use the @smithery/api package to connect programmatically and call Nimble tools from any TypeScript application or AI framework.

​
Install

npm install @smithery/api @ai-sdk/mcp @ai-sdk/anthropic ai

​
Connect and Call Tools

import Smithery from "@smithery/api"
import { createMCPClient } from "@ai-sdk/mcp"
import { generateText } from "ai"
import { anthropic } from "@ai-sdk/anthropic"
import { createConnection } from "@smithery/api/mcp"

const smithery = new Smithery()

// Create a connection with server config
const conn = await smithery.connections.create(
  "{your-namespace}",
  {
    mcpUrl: "https://nimble.run.tools",
    headers: {
      // Authorization Header: "Bearer {NIMBLE_API_KEY}" — required
      "NIMBLE_API_KEY": "Bearer your-NIMBLE_API_KEY"
    },
  }
)

// Connect using the returned connection ID
const { transport } = await createConnection({
  client: smithery,
  namespace: "{your-namespace}",
  connectionId: conn.connectionId,
})

const mcpClient = await createMCPClient({ transport })
const tools = await mcpClient.tools()

const { text } = await generateText({
  model: anthropic("claude-sonnet-4-20250514"),
  tools,
  prompt: "Use the available tools to help me.",
})

await mcpClient.close()

​
Available Tools

Smithery auto-discovers all tools from the Nimble MCP server:
ToolDescription
searchWeb search with configurable depth and focus modes
extractExtract content from any URL with rendering support
mapDiscover all URLs on a website via sitemaps and link crawling
crawlCrawl multiple pages with path filtering and progress tracking
agentsRun pre-built extraction agents for structured data
See the Nimble MCP Server docs for the full list of available tools and configuration options.

​
Next Steps

Nimble MCP Server

Full MCP server setup for Cursor, Claude Desktop, and other clients

Search

Web search with depth levels, filtering, and AI answers

Google ADK

Use Nimble with Google ADK agents via MCP

OpenAI

Use Nimble with OpenAI function calling and the Agents SDK