2/20/2026 | USA | technology | ✓ Verified - arxiv.org

Web Verbs: Typed Abstractions for Reliable Task Composition on the Agentic Web

#Web Verbs #Typed abstractions #Agentic web #Large language models #Web actions #API-browser unification #Preconditions #Postconditions #Policy tags #Logging #Task composition #Semantic layer #Reliability #Traceability

📌 Key Takeaways

Web agents currently depend on low‑level, brittle actions like clicks, leading to inefficiency and hard-to-verify tasks.
The authors introduce Web Verbs—a typed abstraction layer that maps site capabilities to semantically documented functions.
Each verb carries preconditions, postconditions, policy tags, and logging, allowing for strict contracts and traceability.
Web Verbs unify API‑based and client‑side (browser) workflows, enabling large language models to produce concise, reliable programs.
Proof‑of‑concept implementations and case studies demonstrate concise, robust execution compared to existing agents.
A roadmap for standardizing Web Verbs at web scale is outlined to promote deployability and trustworthiness.

📖 Full Retelling

WHO: The paper "Web Verbs: Typed Abstractions for Reliable Task Composition on the Agentic Web" is authored by Linxi Jiang, Rui Xi, Zhijie Liu, Shuo Chen, Zhiqiang Lin, and Suman Nath. WHAT: It proposes a web‑scale set of typed, semantically documented functions—Web Verbs—that provide a uniform interface for web actions, unifying API‑based and browser‑based workflows. WHERE: The manuscript was submitted to arXiv, a preprint repository for scientific research. WHEN: The first version was posted on 19 February 2026. WHY: The authors argue that current web agents rely on brittle low‑level primitives (clicks and keystrokes), which hampers efficiency and verifiability; Web Verbs aim to offer stable, composable units that enable large language models to synthesize reliable and auditable workflows for the agentic web.

🏷️ Themes

Artificial intelligence, Web automation, Large language models, Software agents, Functional abstraction, Reliability and verifiability, Standardization of web interfaces

Entity Intersection Graph

No entity connections available yet for this article.

Deep Analysis

Why It Matters

Web Verbs introduces a typed, semantically documented interface for web actions, enabling reliable and auditable task composition by large language models. This abstraction bridges API and browser paradigms, reducing brittle low-level interactions and improving efficiency for software agents.

Context & Background

Shift from human browsing to agentic web interactions
Current agents rely on brittle clicks and keystrokes
Web Verbs proposes typed function calls with pre/post conditions

What Happens Next

The authors plan to standardize Web Verbs for broader adoption, encouraging sites to expose typed actions. Future work includes expanding the verb library, integrating policy and logging, and evaluating performance in real-world agent deployments.

Frequently Asked Questions

What are Web Verbs?

Typed, semantically documented functions that expose site capabilities through a uniform interface, usable via APIs or browser workflows.

How do they improve agent reliability?

By providing stable interfaces, reducing steps, and enabling typed contracts and traceable execution, agents can compose tasks more efficiently and with verifiable correctness.

}

Original Source

              --> Computer Science > Artificial Intelligence arXiv:2602.17245 [Submitted on 19 Feb 2026] Title: Web Verbs: Typed Abstractions for Reliable Task Composition on the Agentic Web Authors: Linxi Jiang , Rui Xi , Zhijie Liu , Shuo Chen , Zhiqiang Lin , Suman Nath View a PDF of the paper titled Web Verbs: Typed Abstractions for Reliable Task Composition on the Agentic Web, by Linxi Jiang and 5 other authors View PDF HTML Abstract: The Web is evolving from a medium that humans browse to an environment where software agents act on behalf of users. Advances in large language models make natural language a practical interface for goal-directed tasks, yet most current web agents operate on low-level primitives such as clicks and keystrokes. These operations are brittle, inefficient, and difficult to verify. Complementing content-oriented efforts such as NLWeb's semantic layer for retrieval, we argue that the agentic web also requires a semantic layer for web actions. We propose \textbf{Web Verbs}, a web-scale set of typed, semantically documented functions that expose site capabilities through a uniform interface, whether implemented through APIs or robust client-side workflows. These verbs serve as stable and composable units that agents can discover, select, and synthesize into concise programs. This abstraction unifies API-based and browser-based paradigms, enabling LLMs to synthesize reliable and auditable workflows with explicit control and data flow. Verbs can carry preconditions, postconditions, policy tags, and logging support, which improves \textbf by providing stable interfaces, \textbf by reducing dozens of steps into a few function calls, and \textbf through typed contracts and checkable traces. We present our vision, a proof-of-concept implementation, and representative case studies that demonstrate concise and robust execution compared to existing agents. Finally, we outline a roadmap for standardization to make verbs deployable and trustworthy at web scale. Su...
            

Read full article at source

Source

arxiv.org