Meeting Minutes

Meeting Minutes === :::spoiler Document Usage This document is used for all minutes of the WoT CG. After the meeting, the contents will be copied to the GitHub with a PR. === ###### tags: `Meetup` :::spoiler Document Usage This document is used for all minutes of the WoT CG. After the meeting, the contents will be copied to GitHub with a PR. For one week, changes can be requested and then the PR will be merged. After the PR is merged, the content here can be deleted. Document Access: - Everyone has read and commenting access - Cristiano Aguzzi and Ege Korkan can edit the document - Other people can request write access for a meeting - The scribe of the meeting will have write access   ::: :::info :date: **Date:** 24 April 2026 ### :bust_in_silhouette: Participants  - Ege Korkan - Cristiano Aguzzi - Andrei Ciortea - Roman Binkert - Afo - Alexandru Sorici - Andrei Olaru - Antoine Zimmermann - Ben Francis - James Harrison - Jeremy Lemee - Julian Padget - Kunihiko Toumura - Gustavo Nardin - Maud Stiernet - Shakir - Simon Mayer - Tomoaki Mizushima - Vignesh Vaidyanathan :writing_hand: **Scribe:** Cris ---- ### Welcome Ege: Joint call with Web Agents Community Group ### News Ege: Next general WoT Event will happen in China . Ege: TPAC is also planned in Dublin, use the official W3C event to know more. Ege: The main WoT website is getting a new restyle. We will improve the content too. Ege: From WebAgents CG we have built a report about interoperability for Agents on the Web. ### Meetup Presentation Ege: Today Roman Binkert presents his findings about Autonomus agents comunicating in a cyber physical enviroment. Roman: The motivation for this work is the following scenarion. Consider a robot arm that want to move the object. How would it know about the presence of an object? It would need to comunicate a sensor. Roman: The comunication and the capability discovery is handled by WoT. The sensor has a TD and the robot can use it. Roman: When you scale (e.g. adding another robot arm or more objects) you increse the problem of having different interaction modles. Again this can be solved by WoT. Roman: Multi-Agents has autonomy and self-organization. But lacks interoperability. WoT can help in this sense. Roman: The central part of the WoT is the Thing Description. It is use to describe how you can interact with the device. Roman: In a TD you can describe affordances: what a Thing can do in its enviroment Roman: We have properties actions and events. Roman: Agents (a very hot buzzword now) in our usecase we refer to Agents in the context of agentic programming. Agentic programming is another way of programming devices. You don't use regular structured intructions but you describe an high level goal and let the agent figure out how to achive it. Roman: Each agent start from sensors than uses a set of beliefs and desires to operate. Roman: the idea of this work is to match Agent Speak to WoT interaction afforndances. We use TD to describe Agents capabilites. Roman: For mapping: - When ask -> read/observe property or subscribe to event - When achive -> invoke an action - tell write a property or use the "tell" action - askHow is a particular action Roman: We use `@type` in the TD to do mapping between affordances and plans. <Roman showing a real-world example> Roman: That concludes my presentation about how to use WoT in a agent-to-agent scenario to create a protocol agnostic agentic applications. ## Q&A Alexandru: suppose that the red or the blue cube have a particular ontology. The meaning of an operation (affordance) can be described in a ontology or natural languange. How do you alling different ontologies when dealing with inputs or outputs in the TD ? Roman: maybe using LLM we can create this alligment. In our approach we didn't use it. Simon: Agents are different from Things. One strong difference is that an agent might refuse to perform an operation. Agency is the capability of deciding by my own goals. This mapping it seems to override this capability making an Agent effectivly a Thing. Roman: this is something we can definetly improve as future work. The real purpose of the work is the mapping. Simon: this is basically creating special actions that has special meaning. Which is something that we should avoid with WoT. Antonie: what is an agent or a tool? Robotic arm agents? Roman: Each arm has an agent that controls them. JLemee: is not clear how to the mapping. Is the mapping happening at compile time o runtime? Roman: I need some atomic step in scenario, those steps can be re-arranged or even renamed to achive a bigger plan. Ben: you talk about protocol agnostic comunication between agents and things. When it comes when actually activate a particular affordance the agents need to implement all different protocols. Do you think that a new web native protocol would help? Roman: we used node-wot. Our framework just activate an affordance it uses node-wot. Ben: does your agent uses scripting api? Roman: yes. Andrei: It is defdefinitely intresting initiative. We have a similar approach. How the interaction between the agents happens? In particular how these plans are transfered. Is there a possibility to "split" the actions? In other words, if one agent is capable of a complex action one other might need multiple atomic simplier actions. Roman: About you last question, if this the case we need a way to describe it. Andrei: Thanks a lot, nice job. How do you position to approaches with Eclipse LMos or the A2A protocol. Are you looking at eterougenous type of agents? Your agent model is not always applicable. Roman: LLM are difficult to describe them with a TD, we need some other mechanism. Maybe some different abstraction layers. Roman: AskHow is special, but tell is not. To achive interoperability ofc we need to agree in using TD and therefore a similar model. Andrei: Information protocol and hypermedia based interaction for thing description. Cristiano: this special AskHow action could it be ever introduced as a native WoT operation inside the WoT interation model? Roman: in the future we might come up with a new description format that is more capable but for the TD is hard to draw a line. Cristiano: you could have TDs as capabilities maybe. ## :ballot_box_with_check: Resolutions ## :exclamation: Action Items ## :envelope_with_arrow: Feedbacks ## Presentation Summary -