How do I use Paperboy from code or from an AI agent?

Which one should I use?

• An AI agent (Claude Code, Cursor, Codex): add the MCP server, or open the setup page for your client. The agent gets eight tools: list sites, suggest topics, write an article, read it back, check claims, publish, performance and queries.
• A script or your own product: call the REST API. There is an OpenAPI 3.1 file at /api/v1/openapi.json.
• A terminal or CI: use the CLI. The fact check exits non-zero on a failing page, so it works as a CI gate.

How does billing work?

Reading is free: sites, articles, performance and queries cost nothing. Spending calls use credits: a pack is $20 for 100. There is no seat limit and no site limit.

What happens when I ask for an article?

Paperboy picks or takes a topic, reads real sources, writes the article, and runs the fact check. Every cited claim is compared with the saved text of its source. A weak article is held back with the verdict instead of being published. By default the API holds a passed article for your approval too. You publish it with one more call.

Where do I start?

Go to the quickstart. You need an account and an API key from Settings, API keys.

Agents can read all of this as plain text: /llms.txt.