Skip to content

[Interop TF Report] Architectural considerations #92

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
wants to merge 4 commits into
base: main
Choose a base branch
from
Draft

[Interop TF Report] Architectural considerations #92

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
f8c1e6d
Interop report: Add structure for architectural considerations
andreiciortea Aug 29, 2025
4988cef
Refine report structure
andreiciortea Oct 24, 2025
6906ea3
Update terminology
andreiciortea Oct 24, 2025
cacaf9f
Add Rem as co-editor
andreiciortea Oct 25, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
99 changes: 82 additions & 17 deletions TaskForces/Interoperability/Reports/report-interoperability.html
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,11 +22,19 @@
orcid: "0000-0003-0721-4135"
},
{
name: "Your Name",
url: "https://your-site.com"
name: "Rem Collier",
company: "University College Dublin",
url: "https://people.ucd.ie/rem.collier",
orcid: "0000-0003-0319-0797"
}
],
authors: [
{
name: "Jérémy Lemée",
company: "University of St.Gallen",
url: "https://www.alexandria.unisg.ch/entities/person/Jeremy_Lemee",
orcid: "0000-0003-0828-1188"
},
{
name: "Your Name",
url: "https://your-site.com"
Expand Down Expand Up @@ -79,6 +87,15 @@
date: "2000",
href: "https://www.dependability.org/wg10.4/meeting38/10-Lala.pdf",
},
COALA: {
authors: [
"Sumers, Theodore","Yao, Shunyu", "Narasimhan, Karthik", "Griffiths, Thomas"
],
title: "Cognitive architectures for language agents",
date: "2023",
href: "https://openreview.net/pdf?id=1i6ZCvflQJ",
publisher: "Transactions on Machine Learning Research"
},
DALE03: {
authors: [
"Jonathan Dale",
Expand Down Expand Up @@ -181,6 +198,14 @@
href: "https://arxiv.org/abs/2302.07842",
publisher: "Transactions on Machine Learning Research (TMLR)",
},
TOOL: {
authors: [
"Wang, Zhiruo", "Cheng, Zhoujun", "Zhu, Hao", "Fried, Daniel", "Neubig, Graham"
],
title: "What are tools anyway? a survey from the language model perspective",
date: "2024",
href: "arXiv preprint arXiv:2403.15452",
},
WILLMOTT02: {
authors: [ "Steven Willmott" ],
etAl: true,
Expand Down Expand Up @@ -210,24 +235,30 @@ <h2>Introduction</h2>
<h2>Terminology</h2>

<dl>
<dt><dfn id="dfn-agent">Agent</dfn></dt>
<dt><dfn id="dfn-agent">Agent</dfn> or <dfn id="dfn-agent">Autonomous Agent</dfn></dt>
<dd>An entity <a href="#dfn-situated">situated</a> in an environment that perceives its environment and acts on it, over time, in pursuit of its goals. For a detailed discussion of agent definitions, see [[FRANKLIN96]].</dd>

<dt><dfn id="dfn-aip">Agent Interaction Protocol</dfn></dt>
<dd>A specification of communication among two or more <a href="#dfn-agent">agents</a> that states who can say what to whom and when — for example, as message sequence diagrams [[AUML]] or information flows [[BSPL]].</dd>

<dt><dfn id="dfn-artifact">Artifact</dfn> or <dfn id="dfn-tool">Tool</dfn></dt>
<dd>A <a href="https://www.w3.org/TR/webarch/#def-resource">resource</a> [[WEBARCH]] that can be shared and used by <a href="#dfn-agent">agents</a> to support their activities. In some <a href="#dfn-mas">multi-agent systems</a>, agents can construct artifacts to instrument their environments [[JACAMO]].</dd>

<dt><dfn id="dfn-augmented-llm">Augmented Language Model</dfn></dt>
<dd>A language model augmented with abilities such as reasoning, tool use, information retrieval, or storing context across interactions. Unlike an <a href="#dfn-agent">agent</a>, an augmented language model does not actively pursue goals and is not <a href="#dfn-situated">situated</a> in an environment. See also [[TMLR23]] and [[ANTHROPIC24]].</dd>

<dt><dfn id="dfn-llm-agent">LLM Agent</dfn> or <dfn id="dfn-language-agent">Language Agent</dfn></dt>
<dd>An <a href="#dfn-agent">agent</a> that relies on an LLM to guide their internal processes and interactions with the environment, while maintaining control over how they accomplish tasks [[ANTHROPIC24]][[COALA]]. [This is the sort of agent people think about when they talk about Agentic AI.]</dd>

<dt><dfn id="dfn-mas">Multi-Agent System (MAS)</dfn></dt>
<dd>A system composed of <a href="#dfn-agent">agents</a> that are situated in a shared environment and interact with one another to achieve individual or collective goals. Agents can work in collaboration, cooperation, and/or competition. A MAS can be either an open or a closed system. This report is primarily concerned with open MAS.</dd>

<dt><dfn id="dfn-situated">Situatedness</dfn></dt>
<dd>The ability of an <a href="#dfn-agent">agent</a> to interact with its environment directly through perception and action, and to respond in a timely fashion to sensory input.</dd>

<dt><dfn id="dfn-tool">Tool</dfn> or <dfn id="dfn-artifact">Artifact</dfn></dt>
<dd>An instrument that can be shared and used by <a href="#dfn-agent">agents</a> to support their activities. In some <a href="#dfn-mas">multi-agent systems</a>, agents construct artifacts to instrument their environments [[JACAMO]]. In the context of agentic AI, a tool is a functional interface to a program that a language model can invoke. Tools extend the capabilities of LLMs by enabling them to retrieve knowledge not seen during training, perform complex computations, mitigate hallucinations, and perceive or act in an environment [[TOOL]].</dd>

<dt><dfn id="dfn-web-tool">Web-based Tool</dfn> or <dfn id="dfn-web-artifact">Web-based Artifact</dfn></dt>
<dd>A <a href="#dfn-tool">tool</a> or <a href="#dfn-artifact">artifact</a> represented as a <a href="https://www.w3.org/TR/webarch/#def-resource">resource</a> [[WEBARCH]] and accessible through the Web. Such tools may expose interfaces over Web or non-Web protocols—for example, a weather service exposing an HTTP API, a lamp exposing a CoAP API, or a telemetry service exposing an MQTT API. Non-Web protocols can be encapsulated behind hypermedia controls published in a description accessible through the Web, such as a W3C Web of Things (WoT) Thing Description [[wot-thing-description11]].</dd>

<dt>[Term]</dt>
<dd>[To be added]</dd>
</dl>
Expand All @@ -251,7 +282,7 @@ <h3>Visions of Agents on the Web</h3>
<p>Recent years have brought renewed interest in Web-based MAS — as evidenced by the <a href="https://www.dagstuhl.de/21072">Dagstuhl Seminar 21072</a> (Feb. 2021) and <a href="https://dagstuhl.de/23081">Dagstuhl Seminar 23081</a> (Feb. 2023) on "Agents on the Web", which led to the creation of the <a href="https://www.w3.org/community/webagents/">W3C Autonomous Agents on the Web (WebAgents) Community Group</a>. A key enabler for this renewed interest is the Web of Things, which provides new practical use cases for Web agents and realizes several visionary ideas anticipated in the original Semantic Web paper [[SEMWEB01]]. Another key enabler is the recent progress in LLM-based agents that can follow instructions and use tools: just like previous generations of agents, LLM-based agents are designed for specific tasks, underscoring the need for open networks in which agents complement one another's abilities to solve more complex problems. New protocols and frameworks are emerging to support LLM-based agents to discover and use tools, or to discover and interact with other agents — many of them explicitly building on Web standards to foster interoperability (e.g., see the <a href="https://modelcontextprotocol.io/">Model Context Protocol</a>, <a href="https://google-a2a.github.io/A2A/">Agent2Agent Protocol</a>, <a href="https://agent-network-protocol.com/">Agent Network Protocol</a>, <a href="https://eclipse.dev/lmos/">Eclipse LMOS</a>).</p>
</section>
<section data-dfn-for="Foo">
<h3>Conceptual Dimensions for Web-based Multi-Agent Systems</h3>
<h3>Conceptual Dimensions</h3>

<p>A <a href="#dfn-mas">multi-agent system</a> (<abbr title="multi-agent system">MAS</abbr>) has several distinguishing features. One key feature is decentralized control, where each <a href="#dfn-agent">agent</a> makes its own decisions and controls its own behavior — yet the MAS as a whole exhibits coordinated behavior to achieve system-level design objectives. Another key feature is that capabilities, knowledge, and resources are distributed among agents, which creates inter-dependencies: agents participate in a MAS because they need to interact with one another to solve problems that would otherwise exceed their individual capacities. Without such inter-dependencies, the MAS would be a collection of isolated agents — and would not constitute a system at all.</p>

Expand Down Expand Up @@ -281,6 +312,39 @@ <h3>Conceptual Dimensions for Web-based Multi-Agent Systems</h3>
Throughout this report, we use these four conceptual dimensions to organize the discussion and emerging technologies.

</section>
<section data-dfn-for="Foo">
<h3>Architectural Considerations</h3>

<aside class="issue">
<p>Different approaches choose different distribution styles, e.g. MCP uses an RPC-style of distribution. This section discusses these different styles with reference to the Web architecture. The discussion could go into many directions, what emerged in our meetings is a need to look deeper than the agent level of abstraction. What seems important is to identify alignments and misalignments with the Web architecture.</p>
</aside>

<section>
<h3>Design Goals</h3>

<aside class="issue">
<p>MAS are a class of systems known for NFPs such as <strong>flexibility/adaptability</strong>, <strong>openness</strong>, <strong>robustness</strong>, and <strong>scalability</strong> [Weyns, 2010]. The Web is a distributed hypermedia system specifcally designed to meet a well-defined set of NFPs, among which <strong>scalability</strong>, <strong>heterogeneity</strong>, <strong>evolvability</strong>, and <strong>extensibility</strong> [Fielding and Taylor, 2002; Taylor et al, 2010]. Inhereting the properties of the Web in Web-based MAS is desirable, which means we need to design MAS that are aligned with the Web Arch [[HMAS]].</p>
</aside>

</section>

<section data-dfn-for="Foo">
<h3>Architectural Patterns</h3>

<aside class="issue">
<p>Components, connectors, and data view of Web-based MAS. See also Weyns' architectural patterns for situated MAS [Weyns, 2010] and the work on hMAS [Ciortea et al, 2018].</p>
</aside>

</section>

<section data-dfn-for="Foo">
<h3>Considerations</h3>

<aside class="issue">
<p>The main considerations we draw from the above. The focus is on alignment with the design of the Web.</p>
</aside>
</section>
</section>
<section data-dfn-for="Foo">
<h3>State of Web-based Multi-Agent Systems</h3>

Expand Down Expand Up @@ -435,13 +499,6 @@ <h3>Agentic AI</h3>
</section>

</section>
<section data-dfn-for="Foo">
<h3>Architectural Considerations</h3>

<aside class="issue">
<p>Different approaches choose different distribution styles, e.g. MCP uses an RPC-style of distribution. This section discusses these different styles with reference to the Web architecture. The discussion could go into many directions, what emerged in our meetings is a need to look deeper than the agent level of abstraction. What seems important is to identify alignments and misalignments with the Web architecture.</p>
</aside>
</section>
</section>
<section data-dfn-for="Foo">
<h2>Identification</h2>
Expand Down Expand Up @@ -589,14 +646,16 @@ <h3>Discussion</h3>
</section>
</section>
<section>
<h2>Norms, Policies, and Organizations</h2>
<h2>Policies, Norms, and Accountability</h2>

<aside class="issue">
<p>See <a href="https://github.com/w3c/odrl/issues/112">ODRL Issue 112</a></p>
</aside>

<section data-dfn-for="Foo">
<h3>Relevant Standards and Initiatives</h3>

<p>ODRL, DIDs?</p>

</section>
<section data-dfn-for="Foo">
Expand Down Expand Up @@ -627,8 +686,14 @@ <h3>Discussion</h3>
</section>
</section>
<section>
<h2>Conclusions: A Strategy for Agents on the Web</h2>

<h2>Conclusions: A Roadmap for Agents on the Web</h2>

<section>
<h3>Accountability</h3>

<p>Make a case for accountability; what do we need to enable accountability, e.g. transparency? answerability (building a dialogue)?</p>

</section>
</section>
<section>
<h2>Acknowledgements</h2>
Expand Down