mcp serverWeb capture as a native agent tool

Your agent calls scrape_url and screenshot_url the same way it calls any other tool. No wrapper to write. No browser to manage. Works in Claude, Cursor, and any MCP-compatible host.

Get your free API key Read the MCP docs

50 MB free. No credit card.

what it isByteKit tools, natively in your agent

MCP (Model Context Protocol) is an open standard for connecting AI agents to external tools. When you add the ByteKit MCP server to your agent host, your agent sees ByteKit's capture tools in its tool inventory — the same way it sees any other tool. No extra API client to write. No hand-rolled tool definitions to maintain.

install and configureAdd ByteKit to your agent host

Hosted (HTTP) POST https://api.bytekit.com/mcp

Connect your agent host directly to ByteKit. Authenticate with your API key in the Authorization header — no local process to run.

.mcp host config

{
  "mcpServers": {
    "bytekit": {
      "url": "https://api.bytekit.com/mcp",
      "headers": {
        "Authorization": "Bearer $BYTEKIT_API_KEY" 
      }
    }
  }
}

Local (stdio) npx / CLI binary

Run the server locally and connect over stdio transport — keep the process inside your own runtime.

shell

 npx @bytekit/mcp-server --api-key $BYTEKIT_API_KEY

Or set BYTEKIT_API_KEY in your environment and omit --api-key.

toolsWhat your agent can do

scrape_url

Fetch any public URL and return its content as clean markdown, HTML, links, or images. Supports geo-targeted requests with a two-letter country code.

screenshot_url

Capture a screenshot of any public URL. Returns an image URL and metadata. Supports desktop, mobile, and tablet rendering; full-page or viewport-only capture.

get_result

Retrieve a completed screenshot by ID. Use this after screenshot_url when the capture is queued for async processing.

search_docs list_docs get_doc

Search and browse the ByteKit documentation bundle. Useful when your agent needs to look up parameter details, endpoint behavior, or SDK usage without leaving the tool-calling loop.

why not hand-rollWriting your own tool takes an afternoon. Maintaining it takes forever.

If you build it	You also inherit	ByteKit handles
Tool definition	Schema versioning and prompt noise when the API changes	Tool definitions, kept current with the API
Browser or proxy layer	Fleet updates, sessions, timeouts	Capture, screenshots, and geo-targeted requests
Content cleanup	Boilerplate, nav elements, cookie banners	Markdown tuned for LLM consumption
Anti-bot handling	Proxy rotation, fingerprinting, CAPTCHA solving	Works on protected sites. Bundled on every plan.

Add web capture to your agent in two minutes.

50 MB free. No credit card.

Get your free API key Read the MCP docs