Multi Hop AI Agent Suite - Architecture

Multi Hop AI Agent Suite - Architecture | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Multi Hop AI Agent Suite - Architecture Sharan Kumar Yenugula, Revanth Ch, Venkat Kotipally This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8880566/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Deploying AI agents in enterprise settings demands more than just intelligence it requires predictability, transparency, and tight control over how these agents interact with critical systems. Current approaches to AI agent design often suffer from unpredictable behavior, poor visibility into decision-making processes, and challenges in ensuring that executions can be verified and repeated. These issues make it difficult to trust AI agents in environments where mistakes can have real consequences. We present the Multi-Hop AI Agent Suite, a new approach to managing AI agents that treats execution control as a first-class concern. Our system breaks down complex tasks into distinct steps we call ”hops” each representing a clear transition from one state to another. Think of it as turning an AI agent’s work into a well-defined sequence of checkpoints rather than a mysterious black box. A central orchestration layer keeps track of where we are in the process, enforces rules about what’s allowed, and ensures everything happens in the right order. What makes our approach different is that agents themselves don’t hold onto hidden information between steps. They’re designed as clean functions that take inputs and produce outputs without side effects, which means we can replay their work and get the same results every time. We’ve separated the ”what should happen next” logic from the ”how to actually do it” mechanics, giving us fine-grained control over execution while maintaining a complete audit trail of everything that happens. This isn’t just about making agents smarter it’s about making them reliable enough to trust in production environments where consistency and accountability matter. Artificial Intelligence and Machine Learning Multi-agent systems orchestration Tool-augmented AI AI control plane Enterprise AI architecture RAG Agent orchestration AI safety Full Text Additional Declarations The authors declare no competing interests. Supplementary Files testresults.txt Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8880566","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":591380571,"identity":"3d40df01-c1ad-4a4f-b270-2d265e9caa04","order_by":0,"name":"Sharan Kumar Yenugula","email":"data:image/png;base64,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","orcid":"https://orcid.org/0009-0008-5948-4608","institution":"Marri Laxman Reddy Institute of Technology and Management","correspondingAuthor":true,"prefix":"","firstName":"Sharan","middleName":"Kumar","lastName":"Yenugula","suffix":""},{"id":591380572,"identity":"1caad7fd-61ad-40bf-8789-43635d1c446a","order_by":1,"name":"Revanth Ch","email":"","orcid":"","institution":"Marri Laxman Reddy Institute of Technology and Management","correspondingAuthor":false,"prefix":"","firstName":"Revanth","middleName":"","lastName":"Ch","suffix":""},{"id":591380573,"identity":"ab0e1adf-7977-440f-b384-a457e77359bc","order_by":2,"name":"Venkat Kotipally","email":"","orcid":"","institution":"Marri Laxman Reddy Institute of Technology and Management","correspondingAuthor":false,"prefix":"","firstName":"Venkat","middleName":"","lastName":"Kotipally","suffix":""}],"badges":[],"createdAt":"2026-02-14 14:10:59","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-8880566/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8880566/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103049639,"identity":"be58c4ce-15d4-4f3a-a9c8-3618d2ece391","added_by":"auto","created_at":"2026-02-20 07:44:00","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":749756,"visible":true,"origin":"","legend":"","description":"","filename":"AIrp.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8880566/v1_covered_2b146c9d-a870-4bd6-b145-35a6d4dc037b.pdf"},{"id":102889315,"identity":"a58599e7-d4bf-44ee-be7e-785476f4f0b5","added_by":"auto","created_at":"2026-02-18 04:07:31","extension":"txt","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1288,"visible":true,"origin":"","legend":"","description":"","filename":"testresults.txt","url":"https://assets-eu.researchsquare.com/files/rs-8880566/v1/7158df25d9998a5131b8a2f9.txt"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eMulti Hop AI Agent Suite - Architecture\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Marri Laxman Reddy Institute of Technology and Management","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Multi-agent systems, orchestration, Tool-augmented AI, AI control plane, Enterprise AI architecture, RAG, Agent orchestration, AI safety","lastPublishedDoi":"10.21203/rs.3.rs-8880566/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8880566/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDeploying AI agents in enterprise settings demands more than just intelligence \u0026nbsp;it requires predictability, transparency, and tight control over how these agents interact with critical systems. Current approaches to AI agent design often suffer from unpredictable behavior, poor visibility into decision-making processes, and challenges in ensuring that executions can be verified and repeated. These issues make it difficult to trust AI agents in environments where mistakes can have real consequences.\u003c/p\u003e\n\u003cp\u003eWe present the Multi-Hop AI Agent Suite, a new approach to managing AI agents that treats execution control as a first-class concern. Our system breaks down complex tasks into distinct steps we call ”hops” each representing a clear transition from one state to another. Think of it as turning an AI agent’s work into a well-defined sequence of checkpoints rather than a mysterious black box. A central orchestration layer keeps track of where we are in the process, enforces rules about what’s allowed, and ensures everything happens in the right order. What makes our approach different is that agents themselves don’t hold onto hidden information between steps. They’re designed as clean functions that take inputs and produce outputs without side effects, which means we can replay their work and get the same results every time. We’ve separated the ”what should happen next” logic from the ”how to actually do it” mechanics, giving us fine-grained control over execution while maintaining a complete audit trail of everything that happens. This isn’t just about making agents smarter it’s about making them reliable enough to trust in production environments where consistency and accountability matter.\u003c/p\u003e","manuscriptTitle":"Multi Hop AI Agent Suite - Architecture","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-18 04:07:26","doi":"10.21203/rs.3.rs-8880566/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ca3aecd3-2ad6-4d3e-ade2-f273a43281b6","owner":[],"postedDate":"February 18th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":62935082,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2026-02-18T04:07:26+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-18 04:07:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8880566","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8880566","identity":"rs-8880566","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}