Docs

How do I add the Paperboy MCP server to my AI agent?

Let your agent suggest topics, write fact-checked articles and publish them, with one URL and one API key. The server speaks Streamable HTTP and works with any client that can send an Authorization header.

What do I need?

Server URL: https://paperboyseo.com/mcp. Auth: the same pb_live_ API key as the REST API, sent as a bearer token. Get a key in Settings, API keys. A future alias, https://mcp.paperboyseo.com/mcp, will serve the same server.

Claude Code

claude mcp add --transport http paperboy https://paperboyseo.com/mcp --header "Authorization: Bearer pb_live_..."

Then ask: "List my Paperboy sites and suggest three topics for the first one."

Cursor

Add this to .cursor/mcp.json in your project, or ~/.cursor/mcp.json for every project:

{
  "mcpServers": {
    "paperboy": {
      "url": "https://paperboyseo.com/mcp",
      "headers": { "Authorization": "Bearer pb_live_..." }
    }
  }
}

Codex

Add this to ~/.codex/config.toml and export the key in your shell. Check Codex's own MCP docs if your version names the fields differently.

[mcp_servers.paperboy]
url = "https://paperboyseo.com/mcp"
bearer_token_env_var = "PAPERBOY_API_KEY"

Is there a setup page for my client?

Yes, for 39 of them. Each page has the exact config copied from that client's own docs, the date we read them, three example prompts and the client's quirks: all agent setup pages. Examples: Claude Code, Cursor, Codex, OpenClaw, Windsurf, Cline.

Which tools does it have?

Tool	What it does	Credits
`get_sites`	List your sites	0
`suggest_topics`	Next articles to write, each with the reason	1
`write_article`	Start a fact-checked article. Returns an id at once	12
`get_article`	Status, verdict and sources. Poll it after write_article	0
`check_claims`	Fact check a URL or pasted text	2
`publish_article`	Publish an article that passed	0
`get_performance`	Clicks, impressions and position from Search Console	0
`get_queries`	Top queries for a site or one article	0

write_article defaults to publish: false, so the agent should ask you before anything goes live. Keys can be limited to the read scope for agents that should only look.

How does it handle long jobs?

An article takes a few minutes, longer than most MCP clients wait for one tool call. So write_article returns an article id straight away and the agent calls get_article until the status is published or needs_review.

Is there a local (stdio) server?

Not yet. The remote server is the supported way today. The CLI is coming soon and is not on npm yet, so use the REST API for scripts.

Ready to read your first article?

Paste your address. In under ten minutes you read a finished, fact-checked article written for your site. Free, no card.

Prefer to run it yourself? Self-host it (repo coming soon). Using an agent? Add the MCP server.