research papers indexed in the Hypothesis beta (NeurIPS & ICML 2025)
AI models available with guardrails and access rights
GPU on demand, no setup
in Constructor's 30-criteria review vs every alternative
Constructor Research is one place to run science: ready-to-use compute and HPC, secure storage, reproducible environments, and collaboration for research teams, in academia and industry. Science-focused AI sits on top, so you can search the literature by idea and method, form evidence-grounded hypotheses, and move to experiments without switching tools. And because Research is part of the Constructor platform, it shares accounts and data with Learn, Assess, Groups, and the rest of your stack. One solution, one context, every scenario.
Everything a research team needs to compute, collaborate, and reason over the literature, without stitching tools together
Cloud, HPC, and GPU that are ready to use, with no bash-and-ssh setup and no waiting on IT. Scale up when an experiment needs it, scale down when it doesn't.
Ready-to-use templates and customizable workflows take the repetitive setup out of experiments, so runs are consistent and faster to start.
Share access and pin environments so collaborators get the same result on the same data. Versioning across experiments and generated content keeps work traceable.
Share data, algorithms, and results with your team and external partners for immediate feedback, with scoped access so collaborators see only what they should.
Coming 2026
A Research Assistant that runs quick checks against existing data and computational methods, and helps plan and troubleshoot experiments, tuned for science rather than generic code completion.
Beta
Search papers by idea, method, claim, and result, not just keywords. Constructor Hypothesis turns the literature into a knowledge graph you can navigate to find evidence, contradictions, and gaps.
Coming 2026
Work with the tools you already use, Overleaf, Mathematica, Matlab, GitHub, and reference libraries like Zotero, so Research fits your workflow instead of replacing it.
Coming 2026
Call research and literature search from MCP-compatible tools (Claude, Cursor, ChatGPT, or your own agents), so discovery becomes part of daily work instead of a separate step.
Access controls and encryption protect sensitive work. For IP-concerned teams, local indexing and text generation can run inside your own cluster.
From a question to a defensible result, on one platform
Compute, HPC, and GPU ready in the cloud, no manual server or software setup.
Connect datasets, code, and reference libraries. Index your own papers alongside the built-in corpus.
Search the literature by method and claim with Hypothesis, then draft evidence-grounded hypotheses.
Use templates and workflows on scalable compute, with a science-focused AI assistant to help.
Share access, results, and pinned environments so teammates reproduce the same outcome.
Get validation plans, track versions, and export traceable artifacts for review and publication.
Available now
2026
2026 and beyond
Notebook tools and national HPC portals give researchers access. Constructor Research adds science-focused AI, HPC and multi-cloud orchestration, and deployable research apps, with a lower skill barrier and a lower total cost.
| Before | With Constructor Research |
|---|---|
| Server management Bash / ssh setup with IT | Server management Ready to use in the cloud |
| Computational resources Manually handle HPC and GPUs | Computational resources On-demand, with no setup |
| Environment management Manual software install and versioning | Environment management Pre-configured and automatic |
| Team collaboration Complex manual integration | Team collaboration Pre-integrated and ready to use |
| Literature and hypotheses Keyword search across scattered PDFs | Literature and hypotheses Semantic search and a knowledge graph (Hypothesis) |
| Capability | Constructor Research | Google Colab | JupyterHub | Posit Workbench |
|---|---|---|---|---|
| AI research assistant | ✓Research Assistant, 2026 | PartialCode completion | – | PartialPositron Assistant / Copilot |
| Literature exploration & knowledge graph | ✓Constructor Hypothesis | – | – | – |
| MCP / agent-tool integration | ✓2026 | – | – | – |
| Agentic scientific experiment workflows | Coming 2026 | – | – | – |
| Compute & infrastructure | ||||
| On-demand cloud with GPU | ✓ | ✓ | Self-hosted | Self-hosted |
| HPC / external hardware orchestration | ✓ | Managed VMs | Via SSHSpawner | ✓Slurm / K8s launcher |
| Multi-cloud orchestration | ✓ | – | – | – |
| Collaboration & apps | ||||
| Collaborative research environment | ✓Hypothesis project, 2026 | ✓IAM roles | Shared volume | ✓Project ACLs |
| Deployable multi-service research apps | ✓ | – | Hub services | Via Posit Connect |
| Container build & image registry | ✓ | – | No registry | Docker images |
| Platform & deployment | ||||
| Part of a full platform | ✓Compute, AI, collaboration, learning & assessment | – | – | – |
| Deployment options | Cloud, private cloud, on-prem, local indexing for IP-sensitive work | Cloud only | Self-hosted | Self-hosted |
See across a field from a single question. Hypothesis turns papers, code, datasets, and results into a navigable knowledge graph of claims, methods, and evidence, so you move from a question to the most relevant work, the hidden connections, and the unresolved gaps, in one flow.
Bring your own papers
Index your library into the same graph as the built-in corpus and see connections across both.
Fill knowledge gaps
Surface open questions, thin evidence, and contradictions to ground new hypotheses.
Validate ideas and get plans
Pressure-test a hypothesis and get a validation plan with datasets, baselines, and metrics.
Works inside your agents
Reach indexed collections from MCP-compatible tools like Claude, Cursor, and ChatGPT.
"The most valuable signal in science is a quiet contradiction between two results that are both solid. Hypothesis finds those, attributes them to specific claims and papers, and proposes what could explain both."
Prof. Dr. Andrey Ustyuzhanin
Constructor University
Made by researchers, for researchers, across disciplines and across academia and industry.
Physicists, biochemistry and bio-material engineers, nanotechnology and polymer scientists running heavy computation and machine intelligence over shared datasets.
Biologists, genetics and microbiology engineers, biotechnology and bioinformatics scientists who need reproducible pipelines and secure data handling.
Computer scientists, data analysts, and AI, robotics, and cybersecurity engineers who want compute, MCP, and agent workflows without infrastructure overhead.
PIs shaping questions and picking defensible methods, and lab or R&D leads who need a portfolio view of topics, bets, and risks across teams.
Early-career researchers ramping fast on a topic and information specializts standardizing sources, provenance, and retraction awareness.
Enterprise R&D groups that need IP-safe deployment, local indexing inside their own cluster, and results they can hand to the business.
"We use Constructor Research to unify data produced by different research groups, run machine intelligence on the dataset, and accelerate our research as a result."
Konstantin Novoselov Nobel Prize winner
Professor of Physics, National University of Singapore
laureate research groups on the platform
Computational Materials - published with code on Constructor
platform replacing scattered servers, IDEs, and storage
"With Constructor, I have everything all in one: a cluster with resources, an IDE, and Constructor Model to store data and results. I especially like the availability of computational resources and how easy the products are to use."
Dr. Andrei Boiarov
Machine Learning Team Lead, R&D, Constructor Tech
"Sparse representation for machine learning the properties of defects in 2D materials", published in npj Computational Materials, with code and trained model weights hosted on Constructor."
Research team led by K. Novoselov and A. Ustyuzhanin
"Hypothesis maps the literature by revealing claims, contradictions, and methods, helping understanding and accelerating work."
Prof. Dr. Ivar Martin
University of Chicago
Buy Research, and the whole Constructor suite comes with it. Research never arrives alone.
The Research app
Research fits the tools and hardware you already run. Connect your stack, or go deeper with the API and MCP.
Cloud, private cloud, on-premise
HPC and GPU, multi-cloud orchestration
Simulation and advanced compute (HPC, GPU, quantum)
Air-gapped and IP-safe installation
Local indexing and text generation in your cluster
Research data needs protection you can defend, and control you can prove. Your work stays yours.
EU data protection, DPA available
US student-record privacy
Protection for younger learners
Responsible AI practices
Regional hosting options
In transit and at rest
Project-scoped sharing and permissions
Information security [verify]
Constructor Research is one platform for running science: ready-to-use cloud, HPC, and GPU compute, reproducible environments, secure storage, and collaboration, plus science-focused AI for exploring the literature and planning experiments. It is used by academic and industrial research teams across materials science, life sciences, and computer science.
No. Compute, HPC, and GPU are ready to use in the cloud, with environments pre-configured and versioned for you. You can also connect your own hardware, orchestrate across clouds, or deploy on-premise and air-gapped where your work requires it.
Science-focused AI helps you work, it does not replace your judgment. A Research Assistant helps run quick checks and plan experiments, and MCP integration lets you call research and literature search from tools like Claude, Cursor, and ChatGPT. Some of these capabilities are rolling out through 2026.
Hypothesis is Constructor's semantic research navigator. It turns papers, code, and datasets into a knowledge graph of claims, methods, and evidence, so you can search by idea rather than keyword, find contradictions and gaps, and turn a direction into a testable hypothesis with a validation plan. It is the discovery layer that sits alongside the Research compute environment.
Research works with the tools scientists already use, including Overleaf, Mathematica, Matlab, GitHub, and reference libraries like Zotero, with more connectors (Notion, Authorea, SciNote) arriving through 2026. It integrates IDEs such as Jupyter and VS Code, and offers an open API and MCP for custom workflows.
Your data is never used to train models, and you own the artifacts you produce. Work is protected with encryption in transit and at rest, fine-grained permissions, and project-scoped sharing. For IP-sensitive teams, indexing and text generation can run inside your own cluster, with on-premise and air-gapped deployment available.
Join a growing network of resellers, integrators, and solution providers who are bringing Constructor Research to universities, research institutes, and organizations around the world. Whether you're an edtech distributor, a research-infrastructure integrator, an HPC or cloud provider, or a regional training company, we offer flexible partnership models designed to create shared value.
Competitive margins and co-selling support on every deal you bring in.
Training, demo environments, co-branded materials, and a dedicated Partner Content Hub.
Joint campaigns, lead sharing, and visibility in our Partner Locator.
