@inproceedings{ghoshal-etal-2026-scaling,
title = "Scaling Enterprise Cloud Agents to Thousands of {API}s: An Empirical Study of Agent Architecture and {MCP} Tool Boundaries",
author = "Ghoshal, Sandip and
Mrini, Khalil and
Assi, Maram and
Fan, Yadan and
Singh, Jyotika and
Ballesteros, Miguel and
Sun, Weiyi and
Mittal, Anshul and
Benajiba, Yassine and
Bharadwaj, Sujeeth and
Ravi, Sujith and
Roth, Dan",
editor = "Atanasova, Pepa and
Haffari, Gholamreza and
Wang, Minghan",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-industry.179/",
pages = "3040--3055",
ISBN = "979-8-89176-460-6",
abstract = "Enterprise cloud agents increasingly operate over thousands of APIs and SDK operations. Tool use therefore becomes a large-scale classification and grounding problem over API operations, not a simple function-calling task. Earlier production-derived architectures either expose broad operation catalogs to the model or hide selection inside planner agents, causing excessive context growth, repeated discovery calls, and preventable invocation errors. We propose the Operation-Visible Agent Architecture, which keeps operation search, schema inspection, validation feedback, and API invocation visible to a centralized ReAct loop through a deterministic Operation-Resolver MCP. We compare this architecture with lead-worker planning and ReAct-based execution variants to assess how agent orchestration and MCP tool-boundary design affect enterprise-agent performance. On internal cloud-management benchmark tasks spanning 280+ SDK clients and approximately 8K operations, the proposed architecture improves full-success rate from 68{\%} to 77{\%} over the production-derived baseline, while reducing total token usage by 54.1{\%}, median latency by 43.3{\%}, LLM calls by 37.8{\%}, and tool errors by 52.4{\%}. The main lesson is that, at enterprise API scale, tool-boundary design is not an implementation detail: parallel tool calling improves efficiency, but task success improves most when operation resolution remains visible to the main reasoning loop."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ghoshal-etal-2026-scaling">
<titleInfo>
<title>Scaling Enterprise Cloud Agents to Thousands of APIs: An Empirical Study of Agent Architecture and MCP Tool Boundaries</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sandip</namePart>
<namePart type="family">Ghoshal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Khalil</namePart>
<namePart type="family">Mrini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maram</namePart>
<namePart type="family">Assi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yadan</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jyotika</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Miguel</namePart>
<namePart type="family">Ballesteros</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weiyi</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anshul</namePart>
<namePart type="family">Mittal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yassine</namePart>
<namePart type="family">Benajiba</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sujeeth</namePart>
<namePart type="family">Bharadwaj</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sujith</namePart>
<namePart type="family">Ravi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dan</namePart>
<namePart type="family">Roth</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pepa</namePart>
<namePart type="family">Atanasova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gholamreza</namePart>
<namePart type="family">Haffari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Minghan</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-460-6</identifier>
</relatedItem>
<abstract>Enterprise cloud agents increasingly operate over thousands of APIs and SDK operations. Tool use therefore becomes a large-scale classification and grounding problem over API operations, not a simple function-calling task. Earlier production-derived architectures either expose broad operation catalogs to the model or hide selection inside planner agents, causing excessive context growth, repeated discovery calls, and preventable invocation errors. We propose the Operation-Visible Agent Architecture, which keeps operation search, schema inspection, validation feedback, and API invocation visible to a centralized ReAct loop through a deterministic Operation-Resolver MCP. We compare this architecture with lead-worker planning and ReAct-based execution variants to assess how agent orchestration and MCP tool-boundary design affect enterprise-agent performance. On internal cloud-management benchmark tasks spanning 280+ SDK clients and approximately 8K operations, the proposed architecture improves full-success rate from 68% to 77% over the production-derived baseline, while reducing total token usage by 54.1%, median latency by 43.3%, LLM calls by 37.8%, and tool errors by 52.4%. The main lesson is that, at enterprise API scale, tool-boundary design is not an implementation detail: parallel tool calling improves efficiency, but task success improves most when operation resolution remains visible to the main reasoning loop.</abstract>
<identifier type="citekey">ghoshal-etal-2026-scaling</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-industry.179/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>3040</start>
<end>3055</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Scaling Enterprise Cloud Agents to Thousands of APIs: An Empirical Study of Agent Architecture and MCP Tool Boundaries
%A Ghoshal, Sandip
%A Mrini, Khalil
%A Assi, Maram
%A Fan, Yadan
%A Singh, Jyotika
%A Ballesteros, Miguel
%A Sun, Weiyi
%A Mittal, Anshul
%A Benajiba, Yassine
%A Bharadwaj, Sujeeth
%A Ravi, Sujith
%A Roth, Dan
%Y Atanasova, Pepa
%Y Haffari, Gholamreza
%Y Wang, Minghan
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-460-6
%F ghoshal-etal-2026-scaling
%X Enterprise cloud agents increasingly operate over thousands of APIs and SDK operations. Tool use therefore becomes a large-scale classification and grounding problem over API operations, not a simple function-calling task. Earlier production-derived architectures either expose broad operation catalogs to the model or hide selection inside planner agents, causing excessive context growth, repeated discovery calls, and preventable invocation errors. We propose the Operation-Visible Agent Architecture, which keeps operation search, schema inspection, validation feedback, and API invocation visible to a centralized ReAct loop through a deterministic Operation-Resolver MCP. We compare this architecture with lead-worker planning and ReAct-based execution variants to assess how agent orchestration and MCP tool-boundary design affect enterprise-agent performance. On internal cloud-management benchmark tasks spanning 280+ SDK clients and approximately 8K operations, the proposed architecture improves full-success rate from 68% to 77% over the production-derived baseline, while reducing total token usage by 54.1%, median latency by 43.3%, LLM calls by 37.8%, and tool errors by 52.4%. The main lesson is that, at enterprise API scale, tool-boundary design is not an implementation detail: parallel tool calling improves efficiency, but task success improves most when operation resolution remains visible to the main reasoning loop.
%U https://aclanthology.org/2026.emnlp-industry.179/
%P 3040-3055
Markdown (Informal)
[Scaling Enterprise Cloud Agents to Thousands of APIs: An Empirical Study of Agent Architecture and MCP Tool Boundaries](https://aclanthology.org/2026.emnlp-industry.179/) (Ghoshal et al., EMNLP 2026)
ACL
- Sandip Ghoshal, Khalil Mrini, Maram Assi, Yadan Fan, Jyotika Singh, Miguel Ballesteros, Weiyi Sun, Anshul Mittal, Yassine Benajiba, Sujeeth Bharadwaj, Sujith Ravi, and Dan Roth. 2026. Scaling Enterprise Cloud Agents to Thousands of APIs: An Empirical Study of Agent Architecture and MCP Tool Boundaries. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 3040–3055, Budapest, Hungary. Association for Computational Linguistics.