1887

Abstract

Summary

Recent advances in large language models (LLMs) and agentic systems have created new opportunities for automating complex engineering workflows. In subsurface modelling, where simulation pipelines combine structural and geological modelling, petrophysical property estimation, multiphase flow simulation, and history matching, such automation promises to significantly accelerate engineering workflows and decision-making. However, exposing computationally-intensive simulation capabilities to agentic clients requires careful architectural design.

We present the open-darts-MCP, an open-source initiative to develop and evaluate an infrastructure for agent-driven reservoir modelling. We design Model Context Protocol (MCP) servers for the open-darts simulator and investigate three approaches: (1) retrieval-augmented generation (RAG), where the LLM synthesizes simulation scripts from ingested documentation and examples, (2) schema-based generation of constrained input configurations, and (3) explicit API exposure through typed deterministic endpoints. We compare these designs in terms of robustness, reproducibility, and flexibility.

On this basis, we implement an agentic workflow in which agents construct inputs and run simulations through the MCP interface, while validation and supervision remain under user control. The workflow demonstrates how LLM-based agents can automate reservoir simulation tasks within controlled boundaries. Thus, the open-darts-MCP provides a practical step toward reliable agentic workflows for geothermal, C02 storage, and broader reservoir engineering applications.

© EAGE Publications BV
Loading

Article metrics loading...

/content/papers/10.3997/2214-4609.202639027
2026-03-09
2026-02-13

  • From This Site

    /content/papers/10.3997/2214-4609.202639027
    dcterms_title,dcterms_subject,pub_keyword
    -contentType:Journal -contentType:Contributor -contentType:Concept -contentType:Institution
    10
    5
Loading full text...

Full text loading...

References

  1. Altynova, A.Y., Kozhevin, A.A., Dubovik, A.S., Khudorozhkov, R.L., Suurmeyer, N. and Martin, T.J. [2024] Advancing Geoscience with Multi-Modal AI: A Comprehensive Copilot. ADIPEC, D011S020R003.
     [Google Scholar]
  2. Blake, R., Kozman, J. and Pelegrin, L. [2025] Learning from Unstructured Documents: Extracting Value Using Machine Learning and Generative Augmented Intelligence. 2025(1), 1–4.
     [Google Scholar]
  3. Dutra, G., Kakar, K.K., Sijibomioluwa, B. and Gadrbouh, R. [2025] End to End Data Curation and Transformation with a Unique Data Quality Workflow: PVT case study. 2025(1), 1–5.
     [Google Scholar]
  4. Jacinto, M., Rodrigues, T., Oliveira, L.D., Ferraz, Q., Medeiros, G., Medeiros, E., Montalvao, L. and Almeida, R.V.d. [2025] LLM- Driven Smart Agents for Surface Logging: Enhancing Drilling and Geological Intelligence with MCP and A2A Protocols. 2025(1), 1–3.
     [Google Scholar]
  5. Jing, Z., Su, Y., Han, Y., Yuan, B., Xu, H., Liu, C., Chen, K. and Zhang, M. [2025] When Large Language Models Meet Vector Databases: A Survey.
     [Google Scholar]
  6. Lee, Z. and Hall, J. [2025] Leveraging LLMs in Reservoir Simulation Workflows: Opportunities and Pitfalls. 2025(1), 1–5.
     [Google Scholar]
  7. Tempone, P., Casetto, I., Piras, C., Feneri, F., Lamonaca, N., Crottini, A., Orlando, L. and Occhiena, C. [2025] Maximizing Efficiency and Data Quality with a New Proprietary Platform for Discrete well Data Management. 2025(1), 1–5.
     [Google Scholar]
  8. Tibari, I., Giborau, R., Crabie, T., Christ, A., Fekir, A.E. and Bouziat, A. [2025] Enhancing Geoscience Document Mining with Large Language Models through GraphRAG Integration and Agentic Architectures. 2025(1), 1–5.
     [Google Scholar]
  9. Voskov, D., Saifullin, I., Novikov, A., Wapperom, M., Orozco, L., Seabra, G.S., Chen, Y., Khait, M., Lyu, X., Tian, X., de Hoop, S. and Palha, A. [2024] Open Delft Advanced Research Terra Simulator (open-DARTS). Journal of Open Source Software, 9(99), 6737.
     [Google Scholar]
  10. Wiegand, K., Mukundakrishnan, K., Bedewi, M., Kahn, D., Tishechkin, D. and Ananthan, V [2024] ENVOY: An AI assistant for Reservoir Simulation. https://stoneridgetechnology.com/company/blog/envoy-an-ai-assistant-for-reservoir-simulation/. Stone Ridge Technology blog.
     [Google Scholar]
  11. Zhang, S., Xu, H., Jia, Y., Wen, Y., Wang, D., Fu, L., Wang, X. and Zhou, C. [2023] GeoDeepShovel: A platform for building scientific database from geoscience literature with AI assistance. Geoscience Data Journal, 10(4), 519–537.
     [Google Scholar]
/content/papers/10.3997/2214-4609.202639027
Loading
Open-DARTS-MCP: An MCP-Based Architecture for Agentic Reservoir Simulation Workflows
Conference Proceedings 2026, 1 (2026); https://doi.org/10.3997/2214-4609.202639027
/content/papers/10.3997/2214-4609.202639027
/content/papers/10.3997/2214-4609.202639027
Loading

Data & Media loading...

Most Cited This Month Most Cited RSS feed

This is a required field
Please enter a valid email address
Approval was a Success
Invalid data
An Error Occurred
Approval was partially successful, following selected items could not be processed due to error