Skip to content

About

AssetOpsBench - Industry 4.0: A unified benchmark and framework for building, orchestrating, and evaluating domain-specific AI agents for Industry 4.0 asset operations and maintenance, with 460+ scenarios, 5 specialist agents (IoT, FMSR, TSFM, Work Order,...), and multi-agent orchestration blueprints (MetaAgent, AgentHive) over MCP.

Topics

Resources

Contributing

Security policy

Stars

2.4k stars

Watchers

15 watching

Forks

Latest commit

 

History

1,170 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AssetOpsBench

Build. Evaluate. Advance industrial AI agents.

An open ecosystem for industrial AI: data, tools, pipelines, leaderboards, competitions, evaluation science, and a growing research community.

Explore the code · Get the data · Read the paper · Try the tutorial · Leaderboards · Competitions · Eval Science


Research papers¹ Benchmark families MCP servers Competition editions Dataset downloads² Universities represented³
15 6 5 + 1 2 2,654 / last month 4

November 9, 2026 snapshot. ¹ Core research table, including accepted papers and preprints. ² Across three Hugging Face dataset repositories. ³ In the listed academic-project directory. Sources and counting rules.

Industrial intelligence begins with a practical question: can an AI agent turn operational evidence into a reliable next step?

AssetOpsBench brings that question into a reproducible environment. It connects sensor observations, failure knowledge, time-series analysis, and maintenance workflows so researchers and practitioners can build agents, inspect their decisions, and evaluate their performance. The research benchmark includes 141 expert-curated scenarios, with 99 single-agent tasks and 42 tasks requiring multiple agents. Read the benchmark study →

Open data. New platforms. Independent reuse.

Run on Harbor · Explore on Kaggle · Build models and new benchmarks

Verified impact Built on the benchmark family
PHMForge on Harbor · DiagnosticIQ on Kaggle Harbor packages PHMForge as 123 tasks from 99 scenarios, including two battery phrasings. DiagnosticIQ’s public Kaggle release provides 6,690 maintenance questions. Harbor’s data-working setup is separate from the MCP-paper evaluation.
From dataset to fine-tuned models Damarla & Zhu publish FailureSensorIQ retrieval and QLoRA code. Released artifacts include a Qwen3 LoRA adapter and ASTRIS reranker.
Companies building on open releases Samyama / VaidhyaMegha built and evaluated an AOB graph extension and released its derived dataset. Faraday Industrial Benchmark preserves an attributed AOB code/data snapshot; its native results are separate.
Company-reported transfer evaluation Good Start Labs’ Alex Duffy reports testing a Diplomacy-fine-tuned model on AOB. The reported transfer study evaluates industrial tasks after game-based training.
Independent research Ilia Shumailov and Christian Schroeder de Witt co-authored multi-CaMeL’s AOB evaluation. Moharram Challenger co-authored Industrial-Instruction’s FailureSensorIQ experiments.

Beyond benchmark runs: whatmodel and BenchmarkList reuse published DiagnosticIQ results; Mercor maps AOB and PHMForge to economic work.

Sources checked November 10, 2026. Reuse stories and source evidence → · Metric definitions →

Research reach: Cited in IndustryBench, Confucius (Minlan Yu and Soudeh Ghorbani), and FORGE (Dacheng Tao). These are citations, separate from dataset reuse. More research stories and sources →

In the community: Discussed by LG AI Research, featured in the Codabench 2025 newsletter and The AI Report on KDnuggets, with discussions spanning ACL Birds of a Feather, NeurIPS, and AAAI evaluation science.

One ecosystem. Many ways to contribute.

AssetOpsBench ecosystem: benchmarks, tools, pipelines, applications, Eval Science, leaderboards, competitions, research, tutorials, events, university projects, and community coverage.

Explore What you will find Start here
Datasets & benchmarks Operational scenarios, sensor–failure questions, and evaluation resources AssetOpsBench data · FailureSensorIQ data
Tools & integrations Domain tools exposed through MCP for industrial workflows MCP server reference
Pipelines & orchestration Agent execution, planning, and evaluation workflows Setup and execution guide
Papers & research Benchmark design, industrial reasoning, retrieval, planning, and deployed systems Research portfolio
Tutorials & learning Guided labs, notebooks, technical material, and conference tutorials Learning resources
Seminars & events Talks, workshops, demonstrations, and community gatherings Events and challenges
Leaderboards Published comparisons and competition rankings, with their evaluation scope Compare systems
Competitions Public challenges, participation resources, results, and retrospectives Explore challenges
Evaluation science Measurement validity, diagnostics, and evidence for operational claims Benchmarks to evidence
Community voices & coverage Platform spotlights, newsletters, independent commentary, and announcements Around the community
Projects & extensions Research collaborations and community experiments Community projects
Applications & demos Examples of industrial agents and interactive evaluation AssetOpsBench-Live

Quick Start

git clone https://github.com/IBM/AssetOpsBench.git
cd AssetOpsBench
pip install -e .

Follow INSTRUCTIONS.md to configure services and run a documented agent workflow. You can also explore the Hugging Face playground or the Colab notebook.

Active development is on main. Publication-specific code is maintained on branches including IndustryAssetEQA and main-0.x.

From operational evidence to evaluated action

The current repository exposes five domain MCP servers—IoT, failure modes, time-series analysis, work orders, and vibration—plus shared utilities. Its agent implementations include Plan Execute, Deep Agent, Claude Agent, OpenAI Agent, Stirrup, and OpenCode. Explore their interfaces and configuration in the repository and server documentation.

Together, these components support a research workflow:

  1. Choose a task grounded in asset operations.
  2. Connect the evidence through domain tools and data interfaces.
  3. Run an agent or pipeline and record its trajectory.
  4. Evaluate the result against the scenario's expected behavior.
  5. Investigate failures and share improvements others can reproduce.

The KDD hands-on tutorial walks through this process, from MCP tools to a grounded work-order agent and an auditable leaderboard.

A family of industrial AI benchmarks

AssetOpsBench connects complementary research questions across industrial operations. Each benchmark has its own task, release, and evaluation protocol.

Benchmark The question it asks Explore
AssetOpsBench Can agents coordinate tools and evidence to complete industrial operations tasks? Paper · Data
FailureSensorIQ Can models reason about relationships between failure modes and sensors? Paper · Data
DiagnosticIQ Given a triggered symbolic rule, can a model recommend the appropriate maintenance action? Paper · Data on Kaggle
PHMForge Can agents select and sequence algorithm-grounded tools for prognostics and health management? Paper · Data and datasheet
IoTBench / ReActXen Can language agents retrieve the right industrial sensor data? Published study
IndustryAssetEQA Can industrial explanations be grounded in telemetry and failure knowledge? Published study

DiagnosticIQ comprises 6,690 multiple-choice questions derived from 118 expert-authored rule–action pairs across 16 asset types. Its five variants probe distractor pressure, surface changes, prose formulations, condition inversion, and rationales. The October 2026 paper evaluates 29 LLMs and four embedding baselines. DiagnosticIQ v2

PHMForge pairs 99 scenarios across eight asset classes with 39 MCP-native tools in the August 2026 paper. The current dataset card provides separate rotating/aero and lithium-ion configurations and documents overlap with the AssetOpsBench PHM release, so their counts should not be added as independent tasks. PHMForge paper · Dataset card

PHMForge on Harbor: 123 packaged tasks adapt the 99 scenarios for general-purpose agents working from data. This setup omits PHMForge’s MCP servers; its scores are separate from the paper’s evaluation. Package details

The team has announced NeurIPS 2026 acceptance for DiagnosticIQ and PHMForge; PHMForge's dataset card identifies the Evaluations & Datasets track. These are acceptance announcements, distinct from records of conference presentations already delivered.

Open data for industrial intelligence

Resource Research focus Access
AssetOpsBench Task-oriented evaluation of industrial agents; the paper describes a 141-scenario benchmark Dataset · Paper
DiagnosticIQ Rule-to-action maintenance recommendation, with targeted robustness variants Paper · Dataset
PHMForge Industrial prognostics and health management; 99 scenarios in the current dataset card Dataset · Paper
FailureSensorIQ Sensor–failure reasoning: 8,296 multiple-choice questions across 10 industrial asset types Dataset · Code and evaluation
IoTBench / ReActXen Natural-language industrial data access: 400+ tasks across five industrial sites, described in the associated paper Paper · Repository
IndustryAssetEQA resources Telemetry-grounded question answering with a failure-mode knowledge graph; evaluated across four asset types Paper · Code, datasets, and graph
Industrial embedding datasets Nine asset-specific retrieval datasets constructed for the embedding study Published study · Implementation

Use the version that matches your experiment. Dataset repositories evolve: the Hugging Face scenario viewer showed 152 rows when reviewed, while the benchmark paper reports 141 curated scenarios. Record the dataset revision, configuration, and code commit when reporting results. Dataset viewer →

Research impact: datasets in use

Explore the reuse stories and released artifacts, academic extensions, and citation audit. The register distinguishes external experiments, connected research, company reports, and citations, with versions and evidence for each claim.

Research portfolio

The portfolio below brings together AssetOpsBench and related work from its team. Tutorials, bridge contributions, and workshop material are identified separately from research papers.

Benchmarks, operational systems, and reasoning

Venue Work Contribution
NeurIPS 2026 · Acceptance announced DiagnosticIQ: A Benchmark for LLM-Based Industrial Maintenance Action Recommendation from Symbolic Rules Symbolic rule-to-action evaluation; author announcement
NeurIPS 2026 · Evaluations & Datasets, per dataset card PHMForge: Evaluating LLM Agents on Industrial Prognostics through MCP-Native, Algorithm-Grounded Tools Algorithm-grounded PHM tools and scenarios; release record
2026 · Preprint Beyond Static Leaderboards: Predictive Validity for the Evaluation of LLM Agents Synthesis of implementation studies and a proposed measurement framework
2026 · Preprint Results and Retrospective Analysis of the CODS 2025 AssetOpsBench Challenge Competition results and evaluation-methodology analysis
KDD 2026 · Datasets & Benchmarks AssetOpsBench A reproducible environment for industrial agent evaluation; venue acceptance is listed in the project
ACL 2026 · Industry IndustryAssetEQA: A Neurosymbolic Operational Intelligence System for Embodied Question Answering in Industrial Asset Maintenance Connects telemetry representations and failure knowledge for grounded explanations
ICLR 2026 Adaptive Conformal Anomaly Detection with Time Series Foundation Models for Signal Monitoring Adaptive anomaly scores using pretrained forecasting models
AAAI 2026 · Demonstration AssetOpsBench-Live: Privacy-Aware Online Evaluation of Multi-Agent Performance in Industrial Operations Containerized submissions evaluated on hidden scenarios with diagnostic feedback
AAAI 2026 SPIRAL: Symbolic LLM Planning via Grounded and Reflective Search Symbolic planning with grounded and reflective search
IAAI 2026 · Deployed Applications Deployed AI Agents for Industrial Asset Management: CodeReAct Framework for Event Analysis and Work Order Automation Event analysis and work-order automation in operational facilities
IAAI 2026 Diversity Meets Relevancy: Multi-Agent Knowledge Probing for Industry 4.0 Applications Multi-agent generation and refinement of domain-relevant questions
NeurIPS 2025 · Datasets & Benchmarks FailureSensorIQ: A Multi-Choice QA Dataset for Understanding Sensor Relationships and Failure Modes Evaluation of industrial knowledge and sensor–failure reasoning
EMNLP 2025 · Industry ReAct Meets Industrial IoT: Language Agents for Data Access ReActXen agents and IoTBench for industrial data access
EMNLP 2025 · Industry Generalized Embedding Models for Industry 4.0 Applications Retrieval models adapted to industrial asset semantics
EMNLP 2025 · Findings Fine-Tuned Thoughts: Leveraging Chain-of-Thought Reasoning for Industrial Asset Health Monitoring Specialized reasoning for industrial health monitoring

Tutorials, bridge, and workshop contributions

Learn by building

Explore the AAAI 2027 lab. The time-series scientific-discovery lab introduces TSM-MCP-Kit, an sktime-backed MCP server, through evidence inspection, recipe composition, diagnostic refinement, evaluation, and search. Lab materials are being released in stages; consult the site for availability.

Start with a guided lab. The KDD tutorial covers environment setup, industrial MCP tools, agent execution, and evaluation.

Explore the underlying techniques. Study the ReActXen IoT agent tutorial, FailureSensorIQ evaluation code, and AssetOpsBench technical material.

Connect research to implementation. Follow the setup instructions, select a documented pipeline, and retain its configuration alongside your results.

Leaderboards

Find the comparison that matches your question. Model knowledge, agent execution, and competition rankings measure different capabilities.

Comparison What it measures Explore
Kaggle AssetOpsBench leaderboard The page currently describes FailureSensorIQ multiple-choice evaluation across 10 industrial asset types; broader agentic evaluation is described as future work Leaderboard and scope
DiagnosticIQ research leaderboard Maintenance-action recommendation and variant robustness; published tables rather than a verified standalone live service Paper and score definitions
PHMForge research comparisons Framework/model configurations on PHM scenarios; the paper reports a public leaderboard Paper · Dataset release
CODS competition rankings Submitted systems evaluated in planning and execution tracks, including public and hidden evaluation Codabench competition · Results analysis
Reproducible tutorial leaderboard Agent outputs evaluated against static scenario data in a guided workflow KDD tutorial · Evaluation documentation

Read scores alongside the dataset revision, model configuration, track, and scoring protocol. The CODS retrospective shows why: public scores and hidden execution scores had a correlation of −0.13 in its analyzed sample, and changing the scaling of one composite-score term would exchange the top two teams. A ranking is meaningful only with its measurement context.

Competitions

Build an agent. Test it under shared conditions. Learn from the failures. Public challenges turn the ecosystem into a place to experiment, compare approaches, and contribute reproducible evidence.

CODS 2025 · AssetOpsBench-Live

The challenge invited teams to develop industrial agents across sensing, reasoning, and action, including maintenance diagnosis and work-order tasks. Its planning and execution tracks are analyzed in the subsequent retrospective. Official conference challenge page · Codabench platform

The official schedule listed a November 13, 2025 submission deadline and December 20 award ceremony. The later retrospective examines a specific 300-attempt server log and 149 registered teams. These scoped counts should be kept distinct from broader participation totals reported in other project publications.

IJCAI 2026 · Industrial Automation Challenge

Hosted on Kaggle by the organizing team, the challenge explores physics-grounded industrial reasoning. Track 1 evaluates internal model reasoning in a closed-book setting, with maintenance and failure-analysis questions. Its listed submission window was June 15–August 18, 2026. Challenge hub · Track 1 · IJCAI competition directory

Competition research

Results and Retrospective Analysis of the CODS 2025 AssetOpsBench Challenge, a 2026 preprint, studies leaderboard behavior, hidden evaluation, scoring scales, participation, and submitted strategies. It complements the AAAI AssetOpsBench-Live system paper: one describes the platform; the other investigates what the competition actually measured.

Evaluation science: from benchmarks to deployment evidence

What does a benchmark score justify saying about an agent in an operational setting? This question connects benchmark development with measurement research and deployment practice.

Research foundation

Beyond Static Leaderboards: Predictive Validity for the Evaluation of LLM Agents, a June 2026 preprint, synthesizes fourteen parallel implementation studies and argues for evaluating whether rankings transfer beyond their original test conditions. It proposes testable predictive-validity criteria and explicitly notes that existing evidence is insufficient to confirm the position.

AAAI-27 Bridge · From Agent Benchmarks to Deployment Evidence

The Bridge website lists Bridging Evaluation Science and Deployment Practice for February 16–17, 2027, in Montréal, Canada. As of this overview's review date, this is an upcoming activity, with planned outputs.

The program brings benchmark builders and measurement researchers together with workflow owners, domain experts, and practitioners. Its industrial-maintenance case uses a configured agent in the AssetOpsBench simulated environment; a second case addresses software engineering. Participants will derive deployment criteria, inspect evaluation evidence, identify missing failure checks, and specify what a supervised pilot must still establish. Program and cases

Planned outputs include two worked case audits and a versioned protocol draft, building on PREP-Eval and mapping to existing benchmark assessment frameworks. The page lists December 1, 2026 for case or interest statements. Participation and dates

This gives evaluation science a visible place in the ecosystem: connecting scenarios, scoring, diagnostics, and operational claims. The competition retrospective supplies a concrete example of why that connection matters. CODS evaluation analysis

Events and challenges

Public engagement

When Event or activity Explore
AAAI 2027 · Session details pending Hands-on lab on agentic scientific discovery for time series with MCP Project site · University preprint record
February 16–17, 2027 · Upcoming AAAI-27 Bridge: From Agent Benchmarks to Deployment Evidence · Montréal Eval Science program
August 11, 2026 KDD hands-on MCP tutorial · Jeju, South Korea Tutorial and materials
July 7, 2026 ACL BoF: What Makes an Enterprise Agent Benchmark Useful? Tasks, Tools & Trust · San Diego Event page
June 25, 2026 Open Source Responsible AI Meetup: AssetOpsBench lightning talk/demo · St John’s University, New York Meetup agenda
2026 competition edition IJCAI Industrial Automation Challenge: physics-grounded industrial reasoning Official competitions · Challenge site
January 2026 AAAI / IAAI research, demonstration, tutorial, and bridge activities · Singapore Demo program · Tutorial
AAAI 2026 AI4TS keynote by Nianjun Zhou: AssetOpsBench for time series and event-driven industrial operations Workshop program
January 20, 2026 NUS seminar by Dhaval Patel on AssetOpsBench and industrial task automation University event listing
December 4, 2025 NeurIPS social talk: Building Reliable Agentic Benchmarks: Insights from AssetOpsBench · San Diego Conference listing · Talk agenda
October 3, 2025 NJIT ACM workshop: AI Agents and Their Role in Industry 4.0 Applications Project event record
September 2025 CODS AssetOpsBench-Live competition launch Competition · Platform paper

Competition archives are research resources, too. The benchmark study reports 500+ agent submissions to the public competition platform. The IJCAI challenge's Track 1 page lists June 15–August 18, 2026; consult each track's page for its own dates and access details. Benchmark study →

Built with the community

Academic projects & university collaborations

University Listed project areas Explore
Columbia University HPML research projects, pipeline optimization, visual inspection, reusable skills, and PHM Instructor account · Project directory
University at Albany SPIN: structured planning and execution efficiency Paper
Binghamton University Confidence-aware coordination and abstention Project directory
University of Technology Sydney Agentic workflows for naval operations and maintenance Project directory

These names identify affiliations of listed projects; the directory is not an institutional endorsement or a complete count of every academic affiliation in the research portfolio.

AssetOpsBench also provides a foundation for collaborative research. These linked community papers explore complementary directions; their publication status and results should be read in their own sources.

Direction Community paper
Planning efficiency SPIN: Structural LLM Planning via Iterative Navigation for Industrial Tasks
Scenario expansion Synthetic Scenario Generation for Evaluation of Industry 4.0 Agents
Pipeline optimization Evaluating Temporal Semantic Caching and Workflow Optimization in Agentic Plan-Execute Pipelines
Conversational operations Towards Multi-Turn Dialog Systems for Industrial Asset Operations and Maintenance
Knowledge infrastructure Knowledge Graphs as the Missing Data Layer for LLM-Based Industrial Asset Operations

The broader project directory includes work on visual inspection, battery analytics, skills and knowledge plugins, smaller models, time-series performance, and additional industrial domains.

Around the community

AssetOpsBench is appearing in benchmarking communities, newsletters, external directories, and practitioner discussion. These sources show different kinds of engagement: a platform spotlight, a research mention, or an external review each tells its own story.

Platform spotlights and newsletters

  • The AI Report on KDnuggets · January 27, 2026: the issue includes a section on AssetOpsBench and industrial multi-agent coordination. Read the issue.

  • Agent Brief · August 13 and September 9, 2026: discusses AssetOpsBench in the broader move toward domain-specific evaluation and operational evidence. August issue · September issue.

  • Codabench's 2025 year-in-review newsletter includes AssetOpsBench in its industrial-applications spotlight, describing the challenge's coverage from sensing through reasoning to action. Read the newsletter.

  • The June 9, 2025 “Daily Paper” newsletter archive includes the AssetOpsBench paper and a short research summary. This is a newsletter mention; the archive alone does not establish an official Hugging Face endorsement. Read the archived issue.

External coverage and perspectives

Source What it contributes
LG AI Research: AEGIS Dongmin Kim’s March 18, 2026 research blog discusses AOB and its live demonstration, and considers future evaluation on the benchmark
Sharan Harsoor on Medium An October 14, 2026 article focused on AssetOpsBench, industrial tool orchestration, and the distinction between benchmark evidence and deployment claims
Siarhei Krupenich on Medium An August 19, 2025 introductory article about the framework and its industrial-agent motivation
HysenLabs An external review of MCP tools, orchestration, documentation, and adoption considerations
Joshua Berkowitz Commentary on industrial-agent benchmarking and reliability
Industrial Atlas An external dataset-directory entry describing AssetOpsBench's industrial operations focus
Mahal Systems benchmark catalog Practitioner-oriented discussion of the sandbox and architecture comparison, alongside caveats about counts, scoring, and leaderboard status
Lancartech retrospective External commentary on local debugging, score interpretation, and the limits of transferring benchmark results to deployments

These are attributed perspectives, not validation of every claim they contain. Read each review against the release and evaluation protocol it discusses.

Academic collaborators and extension contributors

For named course mentors, project participants, and benchmark co-authors, see the contribution records. External research use and citations are documented separately in the research-use register.

Education and practitioner perspectives

Kaoutar El Maghraoui’s account of industrial mentorship in Columbia’s HPML course describes how AssetOpsBench supplied open research problems for student teams. This is a collaborator’s account of the educational work. Company discussion also includes LG AI Research, EQUA AI, and Lancartech, alongside external reviews such as HysenLabs. These are coverage records, not adoption claims. Separately, EQUA’s autonomy guide and work-order data guide cite AssetOpsBench in practitioner discussions of maintenance AI evidence.

Project news

IBM’s own coverage includes the project introduction, the December 2, 2025 IBM–Kaggle leaderboard announcement, and the January 21, 2026 Hugging Face blog.

Conversations with the benchmarking community

The team also communicates directly through the Codabench community forum:

These posts document team outreach. Community research extensions appear in the project directory above; they provide a further view of how others build on the ecosystem.

See the coverage and events inventory for additional media mentions, conference records, directories, attribution notes, and unresolved leads.

Bring your next industrial AI idea

  • Researchers: evaluate a new reasoning method, retrieval system, planner, or agent architecture.
  • Engineers: explore a documented workflow and contribute a reproducible integration.
  • Educators and students: use the labs as a starting point for a course or research project.
  • Domain experts: help create scenarios that capture operational context and meaningful success criteria.

Explore the contribution guide, browse open issues, or propose a new scenario.


Reviewed November 10, 2026. Headline metrics retain their dated November 9 snapshot. This overview inventories resources identified through the public AssetOpsBench repository and linked primary sources. Paper titles and tracks follow publisher records where available. Repository-only event entries are linked to the project record; public listings may not capture every team activity.


Call for Scenario Contribution

We are expanding AssetOpsBench to cover a broader range of industrial challenges. We invite researchers and practitioners to contribute new scenarios, particularly in:

  • Asset Classes: Turbines, HVAC systems, Pumps, Transformers, CNC Machines, Robotics, Engines
  • Task Domains: Prognostics and Health Management, Remaining Useful Life (RUL) estimation, Root Cause Analysis (RCA), Diagnostic Analysis, Predictive Maintenance

How to contribute:

  1. Define your scenario following our Utterance Guideline and Ground Truth Guideline
  2. Explore the Hugging Face dataset for examples
  3. Submit a Pull Request or open an Issue with the tag new-scenario
  4. Contact us with questions:

Infrastructure Support

Model API access for AssetOpsBench 2.0 is enabled by TokenRouter (PaleBlueDot AI), a unified API platform providing access to leading AI models through a single API endpoint.


Contributors

Thanks to these wonderful people ✨

ShuxinLin
ShuxinLin

💻
DhavalRepo18
DhavalRepo18

💻
ChathurangiShyalika
ChathurangiShyalika

💻
nianjunz
nianjunz

💻
PUSHPAK-JAISWAL
PUSHPAK-JAISWAL

💻
LGDiMaggio
LGDiMaggio

💻
Ayan Das
Ayan Das

💻
jdsheehan
jdsheehan

💻
bradleyjeck
bradleyjeck

💻
jtrayfield
jtrayfield

💻
florenzi002
florenzi002

💻
flamehaven01
flamehaven01

💻
kushwaha001
kushwaha001

💻
Rohith-Kanathur
Rohith-Kanathur

💻
Mohit Gupta
Mohit Gupta

💻
jellyfishing2346
jellyfishing2346

💻
anncli
anncli

💻
ozatamago
ozatamago

💻
tomaspasie
tomaspasie

💻
Sans-Shej
Sans-Shej

💻
sanjaiiv04
sanjaiiv04

💻
SajalGoyla
SajalGoyla

💻
priyamDalmia
priyamDalmia

💻
krishrveera
krishrveera

💻
kmn01
kmn01

💻
Coderlicr
Coderlicr

💻
BrianBai093
BrianBai093

💻
AyalYakobe
AyalYakobe

💻
amaan784
amaan784

💻
Dev-Scodes5
Dev-Scodes5

💻
jack-pfeifer
jack-pfeifer

💻
sandeepkunkunuru
sandeepkunkunuru

💻
srutanik
srutanik

💻
thedgarg31
thedgarg31

💻

Star History

Star History Chart

If AssetOpsBench is useful to your work, please ⭐ star the repo, 🍴 fork it, and tell us what you're building.


MseeP.ai Security Assessment Badge

About

AssetOpsBench - Industry 4.0: A unified benchmark and framework for building, orchestrating, and evaluating domain-specific AI agents for Industry 4.0 asset operations and maintenance, with 460+ scenarios, 5 specialist agents (IoT, FMSR, TSFM, Work Order,...), and multi-agent orchestration blueprints (MetaAgent, AgentHive) over MCP.

Topics

Resources

Contributing

Security policy

Stars

2.4k stars

Watchers

15 watching

Forks

Releases

Packages

Used by

Contributors

Languages