AgentSearch

As AI agents and tools increasingly perform tasks on behalf of users, the digital ecosystem is rapidly filling with dynamic, task-oriented systems that offer overlapping capabilities. This shift raises a fundamental question: how can users and orchestrating systems reliably discover, compare, and select the most appropriate agents for a given task? AgentSearch frames this challenge as a first-class information retrieval problem. Unlike traditional IR, where static documents are retrieved, agent search involves dynamic entities whose suitability depends on capabilities, behavior, reliability, constraints, and task-dependent performance. This workshop brings together researchers and practitioners to develop principled methods for indexing, retrieving, ranking, and evaluating AI agents and tools, establishing a coherent research agenda at the intersection of information retrieval and agentic systems.

The workshop features an interactive format including keynotes, panel discussions, breakout sessions, and poster presentations.

The first workshop on Indexing, Retrieval, and Ranking of AI Agents is accepting submissions that describe original research findings, preliminary research results, proposals for new work, and recent relevant studies already published in high-quality venues.

Submission Types

We welcome various types of submissions:

• Extended Abstracts: A concise summary of work in progress or preliminary findings, outlining the main contribution and approach.
  At most 2 pages (excluding references).
• Research, Perspective, and Demo Papers: We welcome original or recently published research on indexing, retrieval, and ranking of AI agents; perspective papers that suggest future directions and research opportunities; and demo papers that showcase systems for agent search. Submissions may follow the relevant SIGIR track guidelines (short/full papers, perspectives, or demonstrations).
  4-9 pages (excluding references).

Submission Instructions

Page limit does not include references. All content, including any appendix, must fit within the page limit. Note that we care more about the quality of the submission and its potential to spur discussion at the workshop than about page length as long as the paper meets the page limits.

At least one author of each accepted paper must register for the workshop and present the paper in person (strongly preferred) or remotely.

We encourage, but do not require, authors to release any code and/or datasets associated with their paper. For demo papers, we particularly encourage anonymous code submission (e.g., via a link in the anonymized paper).

All submissions must be in PDF format and written in English. Please use the SIGIR 2026 template (ACM two-column format). For LaTeX, the following should be used:
\documentclass[sigconf,natbib=true,anonymous=true]{acmart}

All submissions will be double-blind reviewed by the program committee and judged by their relevance to the workshop, especially to the main themes identified above, and their potential to generate discussion. Submissions should be made through EasyChair: https://easychair.org/conferences/?conf=agentsearch2026, and code should be submitted anonymously. When submitting, select one of the paper types (Extended Abstract, Research, Perspective, or Demo).

Note: This is a non-archival workshop. Accepted papers may be uploaded to arXiv.org, allowing submission elsewhere as they will be considered non-archival. The workshop's website will maintain a link to the arXiv versions of the papers. Authors retain full copyright and are free to submit to other venues.