How do I bypass paywalls responsibly with an AI agent?

Bypass paywalls responsibly with an AI agent by using legitimate access methods only: authenticated sessions, publisher APIs, library/institutional credentials, or the metered free-article allowance most paywalls grant. AgentFetch supports authenticated fetching — pass cookies={"sub_token": "..."} or headers={"Authorization": "Bearer ..."} to fetch_url and it'll send your real subscription credentials. For news, this is the only defensible pattern: pay for NYT/FT/WSJ/Bloomberg/Substack, store the session cookie in a secrets manager, and let the agent fetch as you. For academic content, route through your institution's proxy (most universities provide one). For paywalls that allow N free articles per month, respect the limit — AgentFetch can rotate identity per agent run, but using that to evade metering is fraud. Hard "no-go" patterns: 12ft.io, paywall-stripper extensions, archive.org abuse at scale, Googlebot user-agent spoofing. These violate ToS, get IPs banned, and create legal exposure. The defensible architecture is "fetch tool for public web + auth-aware connectors for paid sources" — use a dedicated MCP server (like mcp-bloomberg-api if it exists, or build one against the publisher's API) instead of HTML scraping behind a paywall.