web bot auth

Wolfgang,
thanks for your words of appreciation for my input, glad it is useful

I am mostly just trying to pull many disparate threads together, and keep
an eye on what is out there, but my capacity to work on details of each
protocol/standard is limited *time, attention etc

At this stage, for me and maybe others, the main challenge is cognitive,
This is why my main task is mapping related efforts, and then align with
the bits that I can understand/make sense of/are relevant to me. This is
roughly the point of AIKR *thanks for joining the CG btw Wolfgang

For example, today I am reading about
https://github.com/cloudflare/web-bot-auth  and here
https://datatracker.ietf.org/wg/webbotauth/about/   and cannot remember
whether this being discussed here and in other related WGs *because the
landscape is vast and so dynamic and our ability to retain and process info
is limited
(and in some cases, definitely shrinking!)

Please let us know if and when we are crossing trajectory with IETF on
agents
and how to synergize

Best regards

Paola Di Maio






On Tue, Apr 7, 2026 at 10:49 PM Wolfgang Wimmer <wwimmer@ssc-slovakia.com>
wrote:

> Dear Paola, Group,
>
> Thank you for such a detailed and generous response.
> Really appreciate you mapping the landscape like this, it helps me a lot,
> I honestly did not know all this
> initatives and I think all of them are somehow different, I also did not
> expect to be the only person working on this.
>
> I have registered the website www.a2wf.org on 17th April last 2025 but
> needed almost a Year to work the details out
> and make it available.
>
> Let me try to respond point by point.
>
> On Q1 and the robots.txt gap:
> Yes, that explainer is useful. The core problem you describe is exactly
> why I created siteai.json.
> robots.txt controls access but says nothing about usage, has no semantic
> layer, and cannot distinguish between
> crawling for training versus real-time retrieval versus agentic
> interaction.
> We see siteai.json as the governance layer that sits above robots.txt. Not
> replacing it, but adding the structured,
> action-level permissions that the current landscape lacks.
> A website operator should be able to say "agents may check product
> availability but may not add items to cart without
> human verification" and that statement needs to be machine-parseable,
> timestamped, and legally referenceable.
>
> On Q2 and interoperability:
> Would very much welcome your contributions. We have been thinking a lot
> about how siteai.json can serve as an interoperability
> bridge across the initiatives you listed. Here is how I see the
> relationships:
>
> AIKR CG and A2WF: Your work on knowledge representation and the MCP Model
> Card Specification addresses how AI systems
> declare their own capabilities and constraints. siteai.json addresses the
> complementary side, how websites
> declare their expectations and rules for agents. Together these form a
> complete capability-policy pair.
> We would welcome AIKR input on the ontological structure of our permission
> model.
>
> AI Agent Protocol CG: Their agent discovery and identity work defines how
> agents find and authenticate each other.
>  A2WF defines what agents may do once they arrive at a website. An agent
> discovered through the AI Agent Protocol would consult siteai.json
> before performing actions.
> We see a clear integration point where the identity requirements in
> siteai.json could reference the DID-based authentication they are
> developing.
>
> WebAgents CG:
> Their Interoperability Report and especially the Organization dimension of
> their conceptual model maps directly to what siteai.json implements,
> governance structures for agent behavior.
> We would benefit from their architectural perspective, and siteai.json
> could serve as a concrete implementation of the governance patterns they
> describe.
>
> WebMCP: This is perhaps A2WF strongest complementary relationship.
> WebMCP declares what tools agents can call on a page. siteai.json declares
> the governance conditions under which those tools may be invoked, rate
> limits, authentication, human verification.
> Together they form a complete capability plus policy stack. As WebMCP
> lands in browsers, the need for a policy layer becomes very important...
>
> VCAP and ATEP: ATEP trust tiers could inform siteai.json identification
> requirements.
> A site might grant higher permissions to agents with verified track
> records.
> VCAP settlement infrastructure could enable paid access tiers defined in
> siteai.json.
>
> agent-card.json: This is the mirror image of siteai.json. Agent self
> description versus site policy declaration.
> We should ensure both formats can cross-reference each other.
>
> llms.txt: Near-perfect complement. llms.txt says "here is what is
> important to read" while siteai.json says: "here are the rules for what you
> may do."
> We are considering adding a field in siteai.json that references the
> site's llms.txt.
>
> ai.txt: The academic DSL proposal with element-level granularity is
> interesting.
> siteai.json operates at the action level rather than the element level,
> but there may be alignment opportunities worth exploring.
>
> On Q3 and coordination:
> We actively welcome coordination with all the groups you mentioned. The
> A2WF W3C Community Group is open to anyone.
> We have also submitted a public comment to NIST NCCoE on agent identity
> and authorization,
> arguing that the website operator perspective needs to be part of the
> governance framework from the start.
>
> Your framing of siteai.json as a formal ontology of website intent
> resonates with us.
> That is exactly how we think about it. Would be very interested in
> exploring a joint session or liaison between AIKR and A2WF
> to see how KR best practices can strengthen the specification.
>
> Looking forward to working together on this.
>
> As this is my first W3C Working group and work, please apologise if I am
> not completely understanding all the formalities which
> are standard or expected in this process.
>
> Best regards
>
> Wolfgang Wimmer
>
>
>
>
> Am 03.04.26 um 14:06 schrieb Paola Di Maio:
>
> Wolfgang
>
> Thank you for the welcome and for initiating this work, congrats for
> making a start.
>
> I have been working on closely related challenges -- including how
> websites and AI agents negotiate capabilities, permissions, and
> identification in machine-consumable ways.
> *On the problem statement (Q1):*
> The gap between what robots.txt can express and what AI agents actually
> need to negotiate is well documented. robots.txt controls access but not
> usage, has no semantic layer, and cannot distinguish between crawling for
> training versus real-time retrieval versus agentic interaction. Proposals
> like ai.txt and llms.txt address parts of this, but a structured JSON
> policy file like siteai.json could provide the richer, machine-parseable
> expressiveness that the current landscape lacks. Btw, do you think is
> explainer is useful?
>  https://github.com/w3c-cg/aikr/blob/main/robot_text_explainer.md
>
> *On use cases and requirements (Q2):*
> I ll be happy to contribute some interoperability dimensions that may
> enrich the spec
>
> *On related efforts to coordinate with (Q3):*
> Several active efforts overlap with A2WF's scope:
>
> W3C AIKR Community Group -- We have published Technical Notes on
> machine-consumable specs and are actively working on WebMCP
> interoperability. Our work on the MCP Model Card Specification addresses
> how AI systems declare their own capabilities and constraints, which is the
> complementary side of what siteai.json addresses.
> W3C AI Agent Protocol Community Group -- Their work on agent discovery,
> identification, and collaboration protocols is directly adjacent.
> W3C Autonomous Agents on the Web (WebAgents) CG -- Focused on Web-based
> multi-agent systems aligned with Web Architecture.
> MCP (Model Context Protocol) ecosystem -- The emerging WebMCP work and
> related IETF drafts (VCAP, ATEP) address protocol-level interoperability
> that siteai.json would need to integrate with.
>
> The IETF well-known URI registration for agent.json (
> github.com/protocol-registries/well-known-uris/issues/66) -- relevant to
> the discovery mechanism for siteai.json.
> The ai.txt proposal and llms.txt convention -- both address subsets of the
> same problem space.
>
> In essence, from AIKR point of view, the proposed 2WF approach aligns with
> knowledge representation in AI because a structured JSON policy file is
> essentially a formal ontology of web-site intent -- it makes implicit human
> expectations about agent behavior explicit, machine-parseable, and
> reasonably inferrable, which is the core role for KR .
>
>
> Look forward to hear others thoughts on everything
>
> Best regards
>
> Paola Di Maio, AI KR CG
>
>
> On Tue, Mar 31, 2026 at 9:14 PM Wolfgang Wimmer <wwimmer@ssc-slovakia.com>
> wrote:
>
>> Hello everyone,
>>
>> As the original proposer of the A2WF Community Group, I would like to put
>> my name forward as Chair.
>>
>> A bit of background: I initiated the A2WF project to address the lack of
>> machine-readable governance for AI agent
>> interactions on websites. The current draft specification is available at
>> https://a2wf.org/specification/
>> and the source repository is at https://github.com/a2wf/spec.
>>
>> As Chair, my priorities would be:
>>
>> - Establishing a regular meeting cadence (likely biweekly calls)
>> - Collecting community input on the draft specification
>> - Coordinating with related efforts at IETF (AIPREF), NIST (CAISI), and
>> other W3C groups
>> - Working toward a first Community Group Report
>>
>> If there are no objections, I would be happy to take on this role. If
>> anyone else is interested in
>> serving as Chair or Co-Chair, please speak up. Shared leadership is very
>> welcome.
>>
>> Best regards,
>>
>> Wolfgang Wimmer
>> --
>>
>>
>> www.SSC-Slovakia.com
>>
>> SSC Sales consulting co.ks.
>> Panonska cesta 47, 851 04 Bratislava, SK
>>
>>
>> *Mobile: +43 676 455 34 85 *
>>
> --
> *Geschäftsführer*
>
> www.SSC-Slovakia.com
>
> SSC Sales consulting co.ks.
> Panonska cesta 47, 851 04 Bratislava, SK
>
> *Mobil: +43 676 455 34 85 Mobil: +42 1 949 369 065*
>