Signed agents: giving user-directed bots their own lane

Bot traffic management has long been a binary exercise: block the malicious scrapers and card testers, allow the transparent crawlers that index and monitor your site. Cloudflare’s verified bots program formalized the latter category, vetting bots that disclose who operates them and what they do.

That model is being strained by the rise of AI agents. Unlike traditional bots, which execute a fixed task for a single operator, agents act on behalf of individual end users — planning trips, comparing prices, or completing purchases. The entity controlling the request is no longer a single company, and that changes how websites should evaluate and treat incoming automated traffic.

To address this, Cloudflare is introducing signed agents, a new classification that extends its verified bots program. The category gives customers a distinct label to use when configuring security rules and appears in Cloudflare’s public bot directory on Radar.

What makes an agent different from a crawler

Cloudflare has been refining bot categories as the ecosystem shifts. In 2023, it grouped AI crawler traffic into an official category; in 2024, when OpenAI launched three distinct bots for search, assistant, and archiving functions, Cloudflare mirrored that with three new categories.

The distinction between those earlier categories and agents is one of intent and control. An AI crawler like GPTBot performs a single repetitive job — fetch and index — on behalf of its operator. An agent, by contrast, is directed by an end user to accomplish a varied task, often through remote browsing infrastructure such as Cloudflare’s Browser Rendering or offerings from Browserbase and Anchor Browser. The intelligence isn’t serving one bot owner’s goal; it’s serving the user who issued the instruction.

How signing works

Signed agents build on Web Bot Auth, the cryptographic method Cloudflare announced in May for verifying bot and agent traffic. HTTP message signatures let bots authenticate themselves so customer origins can identify who is actually making the request.

To qualify as a signed agent, three conditions must be met:

  • The agent is generally directed by an end user rather than a single company or entity.
  • The infrastructure or remote browsing platform the agent uses signs requests via Web Bot Auth, with Cloudflare validating the signatures.
  • The agent complies with Cloudflare’s signed agent policy.

The new classification improves on past frameworks in two ways. First, it adds precision and visibility: the Cloudflare bots and agents directory now includes signed agents alongside verified bots, and bot operators can submit signed agent applications directly from the Cloudflare dashboard. Second, it offers easier controls: Enterprise customers will be able to take action on signed agents as a group in security rules, using a new field that will appear in the dashboard.

Bot owners can apply via the Bot Submission Form in the Cloudflare dashboard, specifying whether the submission targets the signed agents list or the verified bots list. All signed agents are recognized via Web Bot Auth validation.

Verified bots aren’t being replaced. Transparent applications using signed agents can still qualify as verified bots if their service meets the existing policy. Cloudflare Radar’s URL Scanner is one example: it relies on Browser Rendering as a service to scan URLs but qualifies as a verified bot because the bot owner — Cloudflare Radar — directs the traffic and signs each request with a unique Web Bot Auth signature distinct from Browser Rendering’s own.

First cohort and Radar visibility

Browser Rendering has been sending signed Web Bot Auth headers since the launch of Web Bot Auth and consistently receives a bot score of 1 for Bot Management customers. As of today, it appears in the signed agent category.

The first cohort of partner agents classified as signed agents includes ChatGPT agent, Goose from Block, Browserbase, and Anchor Browser. These services are representative of the category because their remote browsers are used by end customers rather than by the companies themselves.

Cloudflare’s public directory on Radar, which launched in 2022 and was expanded in 2025 so every verified bot gets its own dedicated page with traffic patterns, now lists both signed agents and verified bots. The directory is viewable by anyone, whether or not they are Cloudflare customers, and includes observed traffic for each recognized bot and agent.

Current availability and roadmap

Starting today, the bot directory supports both bots and agents, and agent creators can submit applications through their Cloudflare account dashboard. Radar publicly shows signed agents and their traffic patterns. Soon, customers will be able to act on signed agents as a group within firewall rules, mirroring the existing capability for verified bots.

Message signatures give websites a reliable way to know what tools are interacting with them, and give bot builders a path to scale that interaction transparently. Cloudflare positions this as the first step in making agent-website interoperability — or, where desired, non-interoperability — manageable at scale.