DRAFT Agentic Web Working Group Charter
The mission of the Agentic Web Working Group is to develop specifications that facilitate web pages and applications to be leveraged by LLMs and AI agents.
This proposed charter is available on GitHub. Feel free to raise issues.
| Charter Status | See the group status page and detailed change history. |
|---|---|
| Start date | [dd monthname yyyy] (date of the "Call for Participation", when the charter is approved) |
| End date | [dd monthname yyyy] (Start date + 2 years) |
| Chairs |
Heather Flanagan (Spherical Cow Consulting) Liad Yosef (Ora) Yoav Weiss (Shopify) |
| Team Contacts | [team contact name] (0.1 FTE) |
| Meeting Schedule |
Teleconferences: topic-specific calls may be held or something else
Face-to-face: we will meet during the W3C's annual Technical Plenary week; additional face-to-face meetings may be scheduled by consent of the participants, usually no more than 3 per year. See also this group calendar |
Motivation and Background
The Agentic Web is where AI agents interact with the Web on behalf of users, necessitating new tools and standards for managing these interactions effectively.
The Agentic Web Working Group addresses the evolving interaction between AI agents and the Web, providing a framework for web authors to manage these interactions effectively.
Scope
The Agentic Web Working Group explores how AI agents interact with the Web, including through browsers. It develops Web interfaces, data formats, related semantics, and others for exposing and using capabilities, context, and data. Its work includes:
- describing the site capabilities and relevant context for agents
- representing agent and site identity, the party an agent acts for, delegation, and interaction context, including negotiating which identity information agents and sites reveal to each other
- supporting transparency about agent and the Web interactions, including status and lifecycle
- handing off between different agentic interaction modes
- expressing permissions, policies, and constraints on agent actions
- attribution of agent-originated interaction to the page it acts on
- measuring the performance, reliability, and outcomes of agent interactions with web pages and applications
- automation and testing support for the above, such as WebDriver extensions
- a privacy and security threat model for agent interactions with the Web
- guidance for web authors on managing agent interactions
Out of Scope
The following features are out of scope, and will not be addressed by this Working Group.
- Model APIs, inference and execution
- Cryptographic authentication primitives
- Crawler controls
- Agent to Server communications
- Content provenance and disclosures
- Bot detection
Deliverables
Updated document status is available on the group publication status page.
Draft state indicates the state of the deliverable at the time of the charter approval. Choose one: Expected completion indicates when the deliverable is projected to become a Recommendation, or otherwise reach a stable state The Working Group intends to publish the latest state of their work as Candidate Recommendation (with Snapshots) and does not intend to advance their documents to Recommendation.
Normative Specifications
The Working Group will deliver the following W3C normative specifications:
- Site Capabilities for Agents
-
This deliverable defines how web pages and applications can describe the functionality they offer, and relevant context, in a structured form that agents acting on behalf of users can discover and use reliably, rather than having to infer it from the page's presentation. It covers both imperative descriptions, provided through script, and declarative ones, provided through markup, including annotations of existing content such as HTML forms and controls. This includes:
- describing site functionality, for example as tools with a name, a natural-language description, input and output schemas, and hints such as whether an action is read-only, consequential, or requires explicit user acknowledgement;
- providing agents with instructions and reference material for the tasks a page supports, for example as skills scoped to the page;
- how user agents make that functionality available to the agents they host or mediate, such as agents integrated into the browser, agents in browser extensions, and web-based agents running in the page itself or in cross-origin frames, and the lifecycle of its invocation, including results, errors, cancellation, and navigation;
- keeping users informed and in control, including confirmation of consequential actions and handing steps back to the user;
- how site capabilities relate to the page's human-facing interface, including its accessibility;
- the security and privacy model for exposing functionality across origins and frames.
The initial input for this deliverable is the WebMCP specification.
Draft state: WebMCP, adopted from the Web Machine Learning Community Group
Expected completion: [Q1–4 yyyy]
The Working Group's specifications may include extensions to WebDriver and WebDriver BiDi, to enable automated testing of their features and to let automation clients, including agents that drive a browser through WebDriver BiDi, use those features, subject to the same permission and consent requirements as other agents.
Tentative Deliverables
Depending on the incubation progress, interest from multiple implementers, and the consensus of the Group participants, the Working Group may adopt Rec-track specifications addressing the following areas. Where no draft exists yet, the area is listed so that the Working Group can adopt a proposal for it once one has been incubated and meets these criteria:
- Agent Identity
-
Mechanisms for conveying to web pages and applications that an interaction is performed by an agent on a user's behalf rather than by the user directly and, where appropriate, information about the agent and on whose behalf it acts, so that sites can make informed decisions, for example about which actions require the user's direct involvement. This also includes conveying site identity to agents, and letting agents and sites negotiate what each reveals to the other. Such mechanisms are expected to minimize the information they disclose, and to carefully consider their privacy and accessibility implications, including for users who rely on agents as assistive technology. This work does not include detecting agents that do not disclose themselves, nor defining cryptographic authentication primitives; it is expected to build on existing cryptographic building blocks, such as Privacy Pass, and on agent identity established by other means, such as OAuth client authentication or the mechanisms developed by the IETF Web Bot Auth Working Group, rather than duplicate them.
Draft state: No draft. Related requirements have been discussed in the Web Machine Learning Community Group, for example in webmcp#54 (agent identity) and webmcp#365 (attribution of agent-originated events).
- Agent Delegation
-
Mechanisms for users to grant an agent limited, revocable authority to act on their behalf on a site (for example, restricted to specific actions, amounts, or time periods), and for sites to express what authority an action requires and verify that it was granted, including how a site can tell that an action requiring explicit user acknowledgement was taken under delegated authority. This includes carrying a user's authenticated context into a headless or remote agent session, bound to the agent and recognizable by the site as delegated, rather than sharing the user's credentials or session with the agent. This work is expected to build on existing mechanisms, such as Web Authentication, Federated Credential Management, the Digital Credentials API, verifiable credentials, OAuth, and Privacy Pass, rather than duplicate them or define new cryptographic primitives.
Draft state: No draft. Related requirements have been discussed in the Web Machine Learning Community Group, for example in webmcp#44 (action-specific permissions) and webmcp#54 (agents as separately authorized clients).
- Agent Interaction Measurement
-
APIs that enable web pages and applications to measure how agents interact with them in the field, for example: the timing and status of invocations of site capabilities, distinguishing invocation outcomes, prevented or incomplete execution, and missing outcome reports; the timing and composition of an agent's interactions with a page, such as capability invocations, DOM actuation, navigations, and time spent waiting for the user; and coarse outcome signals, such as explicit user feedback, or whether the agent assessed a task as completed. These APIs are expected to integrate with the Performance Timeline and the Reporting API, in coordination with the Web Performance Working Group. They must not expose the content of the user's conversations with the agent, and are expected to leave users in control of whether outcome signals are shared.
Draft state: Proposal in the Web Machine Learning Community Group (webmcp#186)
- Agentic Interaction Modes
-
The different modes in which agents interact with the Web and the transitions between them, how each relates to the Working Group's specifications, and the mechanisms these modes require. Modes differ, for example, in where the agent runs (integrated into the browser, in a browser extension, as web content in a page or in a cross-origin frame, driving a browser remotely through an automation protocol such as WebDriver BiDi, or accessing the Web without a browser), in how it acts on content (through capabilities exposed by the site, through the DOM or the accessibility tree, or through screenshots and synthesized input), and in whether the user is present and able to see and control the session. Mechanisms may include letting web pages know, where appropriate, whether a user can see and control the session (for example, before requesting the user's input), and handing off an interaction that an agent started in a headless or remote user agent to a user agent in front of the user, and back, preserving the relevant state (such as navigation, form state, and in-progress capability invocations), with the user's consent and without exposing that state to parties other than the user and the site.
Draft state: No draft
Other Deliverables
- Agentic Web Threat Model
-
A Group Note analyzing the privacy and security threats that arise when agents interact with the Web on behalf of users, such as prompt injection and misleading capability descriptions or outputs that trick agents, exfiltration of user data, cross-site tracking, inferring sensitive information such as accessibility needs from an agent's behavior, blocking or discriminating against users who rely on agents, and actions the user did not intend, and the mitigations expected from the Working Group's specifications, user agents, agents, and sites. It is expected to follow the W3C Threat Modeling Guide, and to inform the security and privacy considerations of the Working Group's specifications.
Draft state: No draft. The Threat Model for WebMCP is an initial input.
Other non-normative documents may be created such as:
- Use case and requirement documents;
- Test suite and implementation report for the specification;
- Primer or Best Practice documents to support web developers when designing applications.
Timeline
Put here a timeline view of all deliverables.
- Month YYYY: First teleconference
- Month YYYY: First face-to-face meeting
- Month YYYY: Requirements and Use Cases for FooML
- Month YYYY: FPWD for FooML
- Month YYYY: Requirements and Use Cases for BarML
- Month YYYY: FPWD FooML Primer
Success Criteria
Remove this clause if the Group does not intend to move to REC: In order to advance beyond Candidate Recommendation, each normative specification is expected to have at least two independent interoperable implementations of every feature defined in the specification, where interoperability can be verified by passing open test suites. In order to advance beyond Candidate Recommendation, each normative specification must have an open test suite of every feature defined in the specification.
There should be testing plans for each specification, starting from the earliest drafts.
Consider adopting a healthy testing policy, such as: To promote interoperability, all changes made to specifications in Candidate Recommendation or to features that have deployed implementations should have tests. Testing efforts should be conducted via the Web Platform Tests project.
Where the behavior of a feature depends on actions by an agent or the user agent that cannot be triggered from page script (for example, agent-initiated invocation of site capabilities, user confirmation, or agent-originated input), the specification should define the automation hooks needed to test it, for example as WebDriver or WebDriver BiDi extensions, in coordination with the Browser Testing and Tools Working Group.
Each specification should contain separate sections detailing all known security and privacy implications for implementers, Web authors, and users.
For specifications of technologies that directly impact user experience, such as content technologies, as well as protocols and APIs which impact content: Each specification should contain a section on accessibility that describes the benefits and impacts, including ways specification features can be used to address them, and recommendations for maximising accessibility in implementations.
This group is expected to be guided by the following documents:
Consider adding this clause if the Group does not intend to move to REC: All new features should have expressions of interest from at least two potential implementors before being incorporated in the specification.
Coordination
For all specifications, this Working Group will seek horizontal review for accessibility, internationalization, privacy, and security with the relevant Working and Interest Groups, and with the TAG. Invitation for review must be issued during each major standards-track document transition, including FPWD. The Working Group is encouraged to engage collaboratively with the horizontal review groups throughout development of each specification. The Working Group is advised to seek a review at least 3 months before first entering CR and is encouraged to proactively notify the horizontal review groups when major changes occur in a specification following a review.
Additional technical coordination with the following Groups will be made, per the W3C Process Document:
In addition to the above catch-all reference to horizontal review which includes accessibility review, please check with chairs and staff contacts of the Accessible Platform Architectures Working Group to determine if an additional liaison statement with more specific information about concrete review issues is needed in the list below.
W3C Groups
- Accessible Platform Architectures Working Group
- @@ai-accessibility
- Accessible Rich Internet Applications Working Group
- Security Interest Group SING
- Coordination on the Agentic Web Threat Model, building on that group's threat modeling work.
- Web Payment Security Interest Group
- Coordination on agentic payments, including how agent identity and delegation apply to payment authorization.
- Web Machine Learning Community Group
- This group incubated WebMCP, the initial input for Site Capabilities for Agents, and hosts the proposal for Agent Interaction Measurement.
- Web Machine Learning Working Group
- This group develops APIs for machine learning inference in the browser, which are out of scope for this Working Group. The groups will coordinate where agentic features depend on such APIs.
- Verifiable Credentials Working Group
- Verifiable credentials are a possible building block for conveying agent identity and delegated authority.
- Browser Testing and Tools Working Group
- This group develops WebDriver and WebDriver BiDi, for automated testing and for agentic browsing. The groups will coordinate on the WebDriver extensions needed to test this group's specifications, including their use in the web-platform-tests, and to let automation clients use the capabilities defined by this group, as well as on keeping agentic browsing through WebDriver BiDi consistent with this group's specifications.
- Decentralized Identifier Working Group
- Coordination on controlled identifiers and delegation of control, as possible building blocks for agent identity and delegation.
- Federated Identity Working Group
- Coordination on agent identity and delegation mechanisms that may build on Federated Credential Management (FedCM) and the Digital Credentials API.
- Privacy Working Group
- Coordination on the privacy aspects of the Agentic Web Threat Model.
- Technical Architecture Group (TAG)
- The Working Group will help the TAG in areas related to the agentic web that require expertise in this space, for example in reviews of other groups' specifications that affect how agents interact with the Web, or in relevant findings and principles documents.
- Web Application Security Working Group
- Coordination on agent identity, on permissions constraining agent actions, and on the security model for exposing site capabilities to agents, including cross-origin exposure, Permissions Policy, and isolation.
- Web Applications Working Group
- Coordination on the attribution of agent-originated events, to be specified in event specifications maintained by that group, such as UI Events.
- Web Authentication Working Group
- Coordination on agent delegation mechanisms that let users authorize an agent, for example by authenticating with a passkey, without sharing their credentials with it.
- Web Payments Working Group
- Coordination on agent-initiated payments through browser-mediated payment flows, which can keep reusable payment credentials out of the agent's reach.
- Web Performance Working Group
- Coordination on the measurement of agent interactions, which is expected to build on the Performance Timeline, Event Timing, and Reporting API specifications.
- Web Platform Incubator Community Group
- A possible venue for incubating proposals for this group's tentative deliverables.
External Organizations
- IETF Web Bot Auth (webbotauth)
- This group standardizes methods for cryptographically authenticating automated clients, including AI agents acting on behalf of users, to Web sites; such HTTP-level authentication is out of scope for this Working Group. The groups will coordinate on how agent identity established through such mechanisms relates to agent identity as exposed to web pages and applications.
- IETF AI Preferences (aipref)
- This group standardizes building blocks that allow for the expression of preferences about how content is collected and processed for Artificial Intelligence (AI) model development, deployment, and use.
- IETF Web Authorization Protocol (oauth)
- This group develops OAuth, including mechanisms for the authorization of automated agents acting on behalf of users. The groups will coordinate so that agent identity and delegation mechanisms for the Web platform build on, rather than duplicate, this work.
- IETF Privacy Pass (privacypass)
- This group standardizes Privacy Pass, a protocol for issuing and anonymously redeeming unforgeable, unlinkable tokens. The groups will coordinate on its use as a building block for agent identity and delegation.
- IETF JSON Schema (jsonschema)
- This group standardizes JSON Schema, which site capabilities may use to describe their inputs and outputs. The groups will coordinate on its use in web platform APIs.
- Agentic AI Foundation (AAIF)
- The AAIF hosts open agentic AI projects, including the Model Context Protocol (MCP). The groups will coordinate on how this group's specifications relate to MCP, and on related work, such as agent identity and observability.
- Coalition for Content Provenance and Authenticity (C2PA)
- This group develops standards for certifying the source and history of media content, which are out of scope for this Working Group. The groups will coordinate where agents consume or produce content whose provenance matters.
- Decentralized Identity Foundation (DIF) Trusted AI Agents Working Group
- This group develops specifications for the identity, delegated authority, and governance of AI agents, such as KYA-OS. The groups will coordinate on agent identity and delegation.
- FIDO Alliance Agentic Authentication Technical Working Group
- This group extends FIDO2/WebAuthn and digital credentials for AI agents acting on behalf of users, for example with delegated credentials and intent-bound assertions. The groups will coordinate on agent delegation, and on distinguishing user-initiated from agent-initiated actions.
- MCP Apps Working Group
- This group defines the MCP extension for interactive user interfaces, which agent applications render as web content. The groups will coordinate on how such content exposes capabilities to agents, such as tools provided by apps.
- WHATWG
- Coordination on integration with the HTML and DOM standards, such as declarative mechanisms for exposing site capabilities (for example, by annotating forms) and the attribution of agent-originated events.
Participation
To be successful, this Working Group is expected to have 6 or more active participants for its duration, including representatives from the key implementors of this specification, and active Editors and Test Leads for each specification. The Chairs, specification Editors, and Test Leads are expected to contribute half of a working day per week towards the Working Group. There is no minimum requirement for other Participants.
The group encourages questions, comments and issues on its public mailing lists and document repositories, as described in Communication. The group also welcomes non-participants to make technical contributions for ongoing work, provided they agree to the terms of the W3C Patent Policy.
The Chairs should periodically look through the non-W3C-Member contributors to the Working Group or the [name of an associated Community Group, if any] and consider whether each one should be invited to participate as an Invited Expert. If a non-W3C-Member contributor would like to participate in meetings, they are encouraged to apply to be an Invited Expert.
Participants in the group are required (by the W3C Process) to follow the W3C Code of Conduct.
Communication
Technical discussions for this Working Group are conducted in public: the meeting minutes from teleconference and face-to-face meetings will be archived for public review, and technical discussions and issue tracking will be conducted in a manner that can be both read and written to by the general public. Working Drafts and Editor's Drafts of specifications will be developed in public repositories and may permit direct public contribution requests. The meetings themselves are not open to public participation, however.
Information about the group (including details about deliverables, issues, actions, status, participants, and meetings) will be available from the Agentic Web Working Group home page.
Most Agentic Web Working Group teleconferences will focus on discussion of particular specifications, and will be conducted on an as-needed basis.
This group primarily conducts its technical work pick one, or both, as appropriate: on the public mailing list public-[email-list]@w3.org (archive) or on GitHub issues. The public is invited to review, discuss and contribute to this work.
The group may use a Member-only mailing list for administrative purposes and, at the discretion of the Chairs and members of the group, for member-only discussions in special cases when a participant requests such a discussion.
Decision Policy
This group will seek to make decisions through consensus and due process, per the W3C Process Document (section 5.2.1, Consensus). Typically, an editor or other participant makes an initial proposal, which is then refined in discussion with members of the group and other reviewers, and consensus emerges with little formal voting being required.
However, if a decision is necessary for timely progress and consensus is not achieved after careful consideration of the range of views presented, the Chairs may call for a group vote and record a decision along with any objections.
To afford asynchronous decisions and organizational deliberation, any resolution (including publication decisions) taken in a face-to-face meeting or teleconference will be considered provisional. A call for consensus (CfC) will be issued for all resolutions (for example, via email, GitHub issue or web-based survey), with a response period from [pick a duration within:] one week to 10 working days, depending on the chair's evaluation of the group consensus on the issue. If no objections are raised by the end of the response period, the resolution will be considered to have consensus as a resolution of the Working Group.
All decisions made by the group should be considered resolved unless and until new information becomes available or unless reopened at the discretion of the Chairs.
This charter is written in accordance with the W3C Process Document (Section 5.2.3, Deciding by Vote) and includes no voting procedures beyond what the Process Document requires.
Patent Policy
This Working Group operates under the W3C Patent Policy (Version of 15 May 2025). To promote the widest adoption of Web standards, W3C seeks to issue Web specifications that can be implemented, according to this policy, on a Royalty-Free basis. For more information about disclosure obligations for this group, please see the licensing information.
Licensing
This Working Group will use the W3C Software and Document license for all its deliverables.
About this Charter
This charter has been created according to section 3.4 of the Process Document. In the event of a conflict between this document or the provisions of any charter and the W3C Process, the W3C Process shall take precedence.
Charter History
Note:Display this table and update it when appropriate. Requirements for charter extension history are documented in the Charter Guidebook (section 4).
The following table lists details of all changes from the initial charter, per the W3C Process Document (section 4.3, Advisory Committee Review of a Charter):
| Charter Period | Start Date | End Date | Changes |
|---|---|---|---|
| Initial Charter | [dd monthname yyyy] | [dd monthname yyyy] | none |
| Charter Extension | [dd monthname yyyy] | [dd monthname yyyy] | none |
| Rechartered | [dd monthname yyyy] | [dd monthname yyyy] |
[description of change to charter, with link to new deliverable item in charter] Note: use the class |
Change log
Changes to this document are documented in this section.