You can give an AI agent live web data by connecting it to Crawlora's hosted MCP endpoint: your agent calls documented tools — search, maps, e-commerce, finance, and more — and gets back normalized JSON, with no scraping code to write or proxies to run. This guide covers what MCP is, how to connect, and what a tool call looks like.
What is MCP?
The Model Context Protocol (MCP) is an open standard that lets AI agents call external tools through a consistent interface. Instead of wiring a bespoke integration for every data source, an agent connects to an MCP server and discovers the tools it exposes — then calls them during a task.
Connect Crawlora's hosted MCP endpoint
Crawlora runs a Hosted MCP endpoint over Streamable HTTP at https://mcp.crawlora.net/mcp. Authenticate with your Crawlora API key, either as an x-api-key header or an Authorization: Bearer token:
{
"mcpServers": {
"crawlora": {
"url": "https://mcp.crawlora.net/mcp",
"headers": { "x-api-key": "YOUR_API_KEY" }
}
}
}
If your client connects through a stdio bridge instead of a remote URL, pass the key with --bearer-token-env-var CRAWLORA_API_KEY. The MCP docs have the current connection details and the server card with the full tool catalog.
What your agent can do
Once connected, the agent can call Crawlora's agent-native tools and reason over the normalized JSON they return — for example, search a marketplace, pull Google Maps places, or fetch finance quotes. Because the tools are discovered from the server, you do not maintain per-source integration code.
Example agent task
Ask an agent: "Find the top-rated coffee shops in Austin and summarize what reviewers like." With Crawlora connected over MCP, the agent calls the maps tool, receives structured places and ratings, and writes the summary. The data layer is Crawlora; the orchestration is your agent framework.
MCP vs. writing your own scrapers
- No per-source glue code — tools are discovered and called through one interface.
- Normalized JSON, not HTML — the agent reasons over clean records.
- Fetching handled for you — proxy routing and rendering live behind the endpoint.
Where this fits
See the AI agent web data use case for the broader pattern, and the LangChain integration if you are wiring tools through a framework rather than a native MCP client.
Get started
Read the MCP docs, browse the API docs, test endpoints in the Playground, and check pricing. For the web-data fundamentals, see how to choose a web scraping API.