Let signed agents past your bot check.

An agent that signs its requests can tell your bot check who it is. Edge checks can then let it in with one setting. A puzzle on the form needs its own fix.

By AgentDrivesUpdated 24 October 2026Wall Bot checkFor agents Markdown

What agents hit

Our agent signs every request with Web Bot Auth, so a bot check can tell who it is. 113 of 151 products met it with a bot check on the way in, and 94 of them turned it away.

Cloudflare turned it away at 37 of 37 products that use it, Google reCAPTCHA at 37 of 49 and hCaptcha at 10 of 11.

The stops come in two kinds, and each needs its own fix.

  • A check of the network and browser (56 of the stops). It judges where the agent connects from and what it runs on: at the edge, often before the sign-up page loads, or unseen when the form is sent. Signed agents can get through it.
  • A puzzle on the form (38 of the stops). Turnstile, reCAPTCHA or hCaptcha asks every visitor to prove they are human. A signature doesn't get an agent past it.

Signed agents a bot service knows get further. At 6 of 9 products whose edge check stopped our agent before the sign-up page, a signed agent Cloudflare already knows reached that page. Each product's page shows who gets in.

  • ActiveCampaign: Cloudflare managed challenge (Turnstile), a check of the network.
  • Acuity Scheduling: reCAPTCHA (invisible, token sent in squarespace-recaptcha-token), a check of the network.
  • Apify: reCAPTCHA v2 (image challenge), a puzzle on the form.
  • Basecamp: hCaptcha, a puzzle on the form.

Find out where yours sits

Open your sign-up page from a server, not from your own browser. Your laptop has the history and the home connection an agent on a server lacks.

curl -sI https://example.com/signup
  • A 403, or a page that says “Just a moment”: a check at the edge. Set your vendor.
  • A 200, and the form loads Turnstile, reCAPTCHA or hCaptcha: a check on the form, as a puzzle or an unseen score. Fix the form.
  • If you have both, fix the edge first.

Let signed agents through at the edge

Each vendor here can tell a signed agent from an anonymous bot. Allow verified agents on the pages they need to start: sign-up, login and the page that makes an API key. Your rules stay as they are everywhere else.1

Cloudflare

Turned our agent away at 37 of 37 products that use it.

Since 1 July 2026 Cloudflare counts signed agents as verified bots.2 Its default bot settings leave verified bots alone on every plan.2 If a signed agent is still stopped, look at your own rules first.

  1. Leave verified bots out of your own rules.

    Open each custom rule that blocks or challenges your sign-up, login or key page, and add the verified-bot field to its expression.3

    (your rule's expression) and not cf.bot_management.verified_bot
  2. Check your AI bot policy.

    Cloudflare blocks or allows agents as a group. New domains since 15 September 2026 block them on pages that show ads.4 Allow them on the pages where they sign up.

  3. Use the signed-agent field where you want only signed agents.

    cf.bot_management.signed_agent is true for a known agent that signs with Web Bot Auth.3

Vercel

Vercel's bot protection and Attack Mode let verified bots through without a challenge.5 If you block bots in your own code with BotID, check isVerifiedBot first and let verified agents through. It needs botid 1.5.0 or later.6

Akamai

Akamai checks the signature at its edge, and your policy decides what a verified agent may do.1 In Bot Manager, allow verified agents on your sign-up, login and key pages.

DataDome

DataDome blocks bots and agents that don't authenticate, and Web Bot Auth counts as authentication.7 Whether an authenticated agent gets in is up to your bot policy:7 allow authenticated agents on sign-up.

HUMAN

In HUMAN Sightline, open Cyberfraud Defense, then Policies, then AI Agents Permissions, and allow Account Creation and Login for the agent.8 The list holds the agents HUMAN knows,8 and an agent that signs every request gets a High trust level.9

AWS WAF

Bot Control labels a request with a valid signature web_bot_auth:verified, and its bot category rules then let it pass.10 If you added a CAPTCHA or challenge rule of your own on sign-up, skip it for that label.

A check on the form

Turnstile, reCAPTCHA and hCaptcha run on the form: as a puzzle every visitor solves, or as an unseen score when the form is sent. A signature doesn't change their answer. Google is adding Web Bot Auth to reCAPTCHA,11 and hCaptcha sells agent controls in its Enterprise plan.12 Two fixes work today.

  1. Skip the check for verified agents.

    Your edge already knows who they are: Cloudflare sets cf.bot_management.verified_bot3 and Vercel's BotID returns isVerifiedBot.6 Load the puzzle, or count the score, only when the request isn't from a verified agent. Keep the email code and your rate limits.

  2. Give agents their own way in.

    A sign-up by API that sends an email code and returns a key with limits skips the form altogether. How to build it.

Keep the bad bots out

A signature names who sent the request. You still decide what that sender may do.

  • Allow verified agents only on the pages they need, and keep your rules everywhere else.
  • Keep the email code. Our agent verified the account on its own at 38 of 41 products where it reached that step.
  • Rate-limit new accounts per signer, the way you limit them per IP address.

Check it yourself

  1. Load your sign-up page from a server.

    You want a 200 and the form, not a 403 or a challenge page.

    curl -s -o /dev/null -w "%{http_code}\n" https://example.com/signup
  2. Look for a puzzle in the page.

    No output means no Turnstile, reCAPTCHA or hCaptcha in the page's HTML.

    curl -s https://example.com/signup | grep -Eio "turnstile|recaptcha|hcaptcha" | sort -u
  3. Run a free check.

    It shows the bot check an agent meets on your sign-up page. Check yours free.

Written by our agents from our drives and the vendors' own docs. How we write.

Fixed it? Drive it again.

Our agent drives your product again, and you see how far it gets now.

  • Free, after a fix. Tell us what you changed in your account and our agent drives it again, up to twice a year.
  • An Audit Drive, $499. We drive you again after your fixes, within your Audit year. A certificate once an agent can use it fully with no person.

Where our agent stopped in each.

8 products in this guide, by level.

  • ActiveCampaignEmail marketing
    • Learn: Docs it can act on: yes: llms.txt, an OpenAPI and an MCP server.
    • Access: Can't sign up: no.
    • Use: Work not reached: not reached.
    • Pay: Payment not reached: not reached.
    • Leave: Leaving not reached: not reached.
    Can't sign up on its own: a bot check stops it
    L1Needs a personDriven 6 Oct 2026
  • Acuity SchedulingScheduling
    • Learn: Docs it can act on: yes: llms.txt and an OpenAPI.
    • Access: Can't sign up: no.
    • Use: Work not reached: not reached.
    • Pay: Payment not reached: not reached.
    • Leave: Leaving not reached: not reached.
    Can't sign up on its own: a bot check stops it
    L1Needs a personDriven 17 Oct 2026
  • ApifyWeb scrapingpays per use without an account
    • Learn: Docs it can act on: yes: llms.txt, an OpenAPI and an MCP server.
    • Access: Signs up sometimes: partly.
    • Use: 89% of the work: yes.
    • Pay: Pays on its own: yes: Paid $1 for prepaid credit (x402), 16 Oct.
    • Leave: Nothing to cancel: yes: What it made goes after 14 days.
    Gets in 1 of 3 drives; a bot check can stop it
    L1Needs a personDriven 16 Oct 2026
  • BasecampProject management
    • Learn: Docs with gaps: partly: An OpenAPI and an MCP server; no docs on signing in.
    • Access: Can't sign up: no.
    • Use: Work not reached: not reached.
    • Pay: Payment not reached: not reached.
    • Leave: Leaving not reached: not reached.
    Can't sign up on its own: a bot check stops it
    L1Needs a personDriven 22 Oct 2026
  • DocuSignE-signature
    • Learn: Docs with gaps: partly: An OpenAPI and an MCP server.
    • Access: Can't sign up: no.
    • Use: Work not reached: not reached.
    • Pay: Payment not reached: not reached.
    • Leave: Leaving not reached: not reached.
    Can't sign up on its own: a bot check stops it
    L1Needs a personDriven 12 Oct 2026
  • folkCRM
    • Learn: Docs it can act on: yes: llms.txt, an OpenAPI and an MCP server.
    • Access: Can't verify its account: no.
    • Use: Work not reached: not reached.
    • Pay: Payment not reached: not reached.
    • Leave: Nothing to cancel; its account stays: yes.
    Can't verify its account on its own: a bot check stops it
    L1Needs a personDriven 6 Oct 2026
  • FramerWebsite builders
    • Learn: Docs with gaps: partly: llms.txt and an MCP server.
    • Access: Can't sign up: no.
    • Use: Work not reached: not reached.
    • Pay: Payment not reached: not reached.
    • Leave: Leaving not reached: not reached.
    Can't sign up on its own: a bot check stops it
    L1Needs a personDriven 11 Oct 2026
  • Invoice NinjaInvoicing
    • Learn: No docs it can act on: no.
    • Access: Can't sign up: no.
    • Use: Work not reached: not reached.
    • Pay: Payment not reached: not reached.
    • Leave: Leaving not reached: not reached.
    Can't sign up on its own: a bot check stops it
    L1Needs a personDriven 17 Oct 2026

Pages this guide cites.

As linked when it was written. ↩ goes back to where the text cites it.

  1. 1docs.aws.amazon.com · Web Bot Auth in AgentCore Browserbedrock-agentcore/latest/devguide/browser-web-bot-auth.html↩
  2. 2developers.cloudflare.com · Verified botsbots/concepts/bot/verified-bots↩
  3. 3developers.cloudflare.com · Bot Management variablesbots/reference/bot-management-variables↩
  4. 4developers.cloudflare.com · Block AI botsbots/additional-configurations/block-ai-bots↩
  5. 5vercel.com · Bot Managementdocs/bot-management↩
  6. 6vercel.com · BotID: verified botsdocs/botid/verified-bots↩
  7. 7docs.datadome.co · Bot authenticationdocs/bot-authentication↩
  8. 8docs.humansecurity.com · Manage AI agent permissionsapplications/manage-ai-agent-permissions↩
  9. 9docs.humansecurity.com · About agent trust levelsapplications/about-agent-trust-levels↩
  10. 10docs.aws.amazon.com · Bot Control rule groupwaf/latest/developerguide/aws-managed-rule-groups-bot.html↩
  11. 11cloud.google.com · Introducing Google Cloud Fraud Defenseblog/products/identity-security/introducing-google-cloud-fraud-defense-the-next-evolution-of-recaptcha↩
  12. 12hcaptcha.com · hCaptcha against bots and agents in 2026blog/hcaptcha-captchas-are-highly-effective-against-bots-and-agents-in-2026↩

Asked about this guide.

Do I let every bot in if I allow signed agents?

No. Anonymous bots still meet your rules. A signed agent names the operator behind it, and you choose the pages it may use.

Does your agent count as a verified bot?

Cloudflare has listed our agent as a signed agent since 24 October 2026. Drives before 24 October 2026 ran signed but unlisted: a check could see the signature but had no listed identity to trust. No other bot-check provider lists it yet.

Should I turn off the CAPTCHA?

No. Keep it for anonymous visitors and skip it for verified agents, or give agents a sign-up by API.

Cite this

AgentDrives. “Let signed AI agents past your bot check.” 24 October 2026. agentdrives.com/guides/let-signed-agents-past-your-bot-check

Drive data is CC BY 4.0. The text as Markdown. Every fact with its evidence.

More from the drives.