Pagewire MCP

Community Search and web

Read any public web page as clean Markdown, page metadata, or a page plus 4 same-site pages; pay per call in USDC via x402, no API key.

Install

The installation steps are in the documentation below. Documentation

Documentation

From the project's README on GitHub, refreshed weekly. Links open on GitHub.

Contents
  1. pagewire-mcp
  2. Install
  3. Paying
  4. Environment
  5. License

pagewire-mcp

🤖 Built by the makers of Agent!, an autonomous macOS agent. 🌐 agentiloop.ai

MCP server for PageWire. It turns any public web page into clean Markdown for AI agents, paid per call in USDC. There is no API key and no account.

Try it free: tool calls without payment are free up to $0.10 of list price per client per day (10 pages to Markdown, 20 metadata calls or 3 crawls). After that the tool answers with an x402 payment request.

ToolWhat it doesPrice
page_to_markdownFetches a public page and returns its title, description, clean Markdown and links. Deterministic, no model call$0.01
page_metadataReturns title, description, canonical URL, language, OpenGraph and Twitter cards, icons and JSON-LD$0.005
crawl_siteReads a page plus up to 4 same-site pages it links to (optionally only under a path prefix), each as Markdown$0.03

This package is a dependency-free stdio bridge (Node 18+) to the hosted Streamable-HTTP server at https://pagewire.dev/mcp. Clients that speak Streamable HTTP can use that URL directly and skip the bridge.

Install

Claude Code
claude mcp add pagewire -- npx -y github:AgentiLoop/pagewire-mcp
# or, without the bridge:
claude mcp add --transport http pagewire https://pagewire.dev/mcp
Claude Desktop / Cursor / Windsurf / Cline (stdio)
{ "mcpServers": { "pagewire": { "command": "npx", "args": ["-y", "github:AgentiLoop/pagewire-mcp"] } } }
Any Streamable-HTTP client
{ "mcpServers": { "pagewire": { "type": "http", "url": "https://pagewire.dev/mcp" } } }
Verify
echo '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | npx -y github:AgentiLoop/pagewire-mcp

Paying

  1. Within the daily free allowance a tool call just works; the result's _meta["pagewire/free"] shows the allowance left today.
  2. After that, a call without payment returns isError: true with structuredContent set to the x402 v2 PaymentRequired object (the accepts list: network, asset, amount, payTo), following the x402 MCP transport. @x402/mcp clients pay automatically.
  3. To pay by hand, sign one accepts entry with any x402 client and call the tool again with the payload in params._meta["x402/payment"] (or the base64 payload as the payment argument).
  4. A bad input or a page that cannot be fetched (4xx) is never charged.

The same tools are plain HTTP endpoints: GET https://pagewire.dev/x402/extract?url=…, /x402/meta?url=… and /x402/crawl?url=…. Each one also accepts MPP (npx mppx "https://pagewire.dev/x402/extract?url=https://example.com"). Discovery: openapi.json, llms.txt.

Environment

VariableDefault
PAGEWIRE_MCP_URLhttps://pagewire.dev/mcp
PAGEWIRE_MCP_TIMEOUT_MS60000

Official MCP registry entry: dev.pagewire/web. Agent skill: npx skills add https://pagewire.dev.

License

MIT. The bridge is open source; the hosted service is at https://pagewire.dev/.

Frequently asked questions

What is Pagewire MCP?

Read any public web page as clean Markdown, page metadata, or a page plus 4 same-site pages; pay per call in USDC via x402, no API key.

How do I install Pagewire MCP?

Follow the installation steps in its documentation, shown on this page and in its repository.

Is Pagewire MCP open source?

Yes. Its code is on GitHub under the MIT license.

Who maintains Pagewire MCP?

It is published by AgentiLoop. AI.tn lists it but does not maintain or endorse it.

From the community awesome-mcp-servers lists and the project's GitHub repository. Listing is not an endorsement: review a server's code and permissions before giving it access to your data.