Import.io

YOU

Scrape the top five stories from Hacker News and return title, points and link.

Select “Play sequence” to see how the tools cooperate. No network request is made.

10,000 free successful calls

$0.0002 per successful call after

OAuth or bearer API key

Hosted no scraping stack to run

One endpoint. Your client.

Choose the client you already use. The configuration below is real, copyable and paired with the right authentication path.

01 · Choose your client Select a logo to update the setup

02 CLAUDE SETUP

AUTH EXAMPLE

Install the Import.io plugin in Claude, then use hosted web scraping tools directly in the conversation.

Open the complete setup guide →

Browsing finds a page. This returns usable data.

Repeatable extraction needs a browser, geographic control and a schema, not a wall of HTML in the model context.

01 · render

Dynamic pages need a browser.

Run JavaScript, wait for the interface and interact with controls before extracting.

02 · locate

Location changes the answer.

Target the country your workflow needs so price, stock and content reflect the right market.

03 · structure

Raw pages waste context.

Return the requested fields as structured data instead of sending full HTML to the model.

04 · continue

Blocks should not stop the agent.

Proxy routing and CAPTCHA handling sit behind the tool call, not inside your workflow.

Give the agent the right web primitive.

Explore what each tool contributes to the workflow. The example input and result change with the selected tool.

Render the page the way a user sees it.

Use a hosted browser for JavaScript, lazy loading and interactive content before another tool extracts or inspects the result.

Use it when the page depends on JavaScript or browser state.

Your agent asks. Import.io does the web work.

One hosted endpoint sits between the client and a production scraping engine. The result returns in the form the workflow requested.

CLIENT Claude · ChatGPT
Cursor · Codex

→

HOSTED MCP Render · act
extract · inspect

→

RESULT JSON · HTML
screenshots

Pay for results. Not attempts.

Start with 10,000 successful tool calls free. After that, pay $0.0002 per successful call and set a monthly spending limit you control.

$0.0002 per successful tool call

• Successful render or action Billed
• Successful screenshot or extraction Billed
• Malformed or failed upstream request Not billed
• Blocked request Not billed

Proxy routing, country targeting and CAPTCHA support included.

Questions developers ask.

The short version. The complete setup and tool reference lives in the MCP documentation. For direct application calls, explore the Web Data API.

How is this different from built-in browsing?+

Built-in browsing is useful for one-off access. Import.io adds hosted browser execution, proxy routing, country targeting, CAPTCHA handling and structured extraction for repeatable workflows.

How do I connect it?+

Add https://mcp.import.io/mcp to your compatible client. Most clients start OAuth automatically; scripts and clients without OAuth can use a bearer API key.

What counts as a billable call?+

Only successful tool calls. Malformed requests, blocked requests and failed upstream calls are not billed.

Can I control spend?+

Yes. Set a monthly spending limit in billing. Paid usage stops when the configured limit is reached.