[FEATURE] Let Claude pass Cloudflare checks as a verified agent acting on behalf of a user
What I was doing:
I asked Claude Code (Opus 5.5) to log into twitterapi.io using my Google account and retrieve my API key. It's my own paid account, and the service explicitly supports AI agents.
What happened:
Claude signed in successfully but got blocked by Cloudflare Turnstile on the dashboard. It refused to proceed because bypassing the check would mean pretending to be human, even though I had explicitly authorized the task.
A simple one-minute step ended up delaying my work by an hour.
Why it matters:
CAPTCHAs are meant to prevent spam, fraud, and abuse, not block authorized AI agents from accessing their users' own accounts.
Cloudflare already has a Signed Agents program that allows verified agents to identify themselves. ChatGPT agent was among the first participants (Cloudflare blog, August 28, 2025).
Claude should have a similar option.
Request:
- Signed agent support: Register Claude's browser and computer-use traffic through Cloudflare's Web Bot Auth or a similar system, so it can identify itself as an authorized agent without pretending to be human.
- Reconsider CAPTCHA restrictions: Review Claude's current restrictions on passing CAPTCHAs, including the possibility of removing them entirely when the agent is performing legitimate, user-authorized actions. A CAPTCHA shouldn't automatically stop a task just because an AI agent is doing it instead of a human.