SandboxAQ is a US-based enterprise platform built around its own Large Quantitative Models (LQMs), which combine quantum technology and AI. Where large language models such as ChatGPT and Claude learn patterns in language, LQMs draw on the laws of physics and chemistry together with proprietary scientific data to predict how molecules and materials behave in the real world. The company offers several products spanning drug discovery, materials science, medical diagnostics, navigation, and cybersecurity, aimed primarily at research institutions and large enterprises. As of 2026, a partnership with Anthropic has produced an MCP server that lets Claude call LQMs directly, making scientific simulations that previously required specialized computing environments accessible through conversation.
Key Features
- Large Quantitative Models (LQMs): Models trained on numerical data from physics and chemistry rather than on language. They aim to predict properties such as binding strength between molecules or atomic-level material behavior at accuracy close to experiment. The subject matter is fundamentally different from text-generating AI
- Applications in drug discovery and materials design: Supports ranking the potency of protein–ligand pairs and pre-screening catalysts and battery materials. The goal is to narrow candidates before laboratory work, lowering failure rates and shortening development timelines
- Conversational use via an MCP server: An MCP server connects to Claude Desktop, claude.ai, and Claude Code. Two tools — AQPotency (protein–ligand potency evaluation) and AQCat Adsorption Spin (adsorption calculations for heterogeneous catalysts) — can be called without provisioning your own compute infrastructure
- AQtive Guard (security): Posture management that surfaces cryptographic assets, keys, and certificates across an organization, extended in March 2026 with AI Security Posture Management (AI-SPM) capabilities: shadow AI detection, guardrails against prompt injection, MCP server risk analysis, and reporting aligned with frameworks such as the EU AI Act
- AQMed / AQNav: Work in medical devices aimed at cardiac disease detection (AQMed) and GPS-independent magnetic navigation (AQNav), both built on quantum sensing technology
- Built for enterprises and research institutions: Not a consumer chat tool but a product portfolio intended for organizations in pharmaceuticals, chemicals, energy, defense, and finance
Pricing
| Plan | Price | What’s included |
|---|---|---|
| AI Simulation (individual, self-serve) | Usage-based | $2,000 in complimentary credits for new accounts, valid for 30 days. Tools such as AQPotency use tiered per-unit rates |
| AQtive Guard | Contact sales | Contract pricing that scales with the cryptographic assets, keys, and certificates discovered. Also available through AWS Marketplace |
| Enterprise / custom LQM deployment | Contact sales | Individual contracts based on use case, data, and scale |
Pricing is current as of August 2026. Published prices are limited and most products are quoted individually. Check the official site for the latest information.
Pros & Cons
✅ Pros
- Predictions grounded in physics and chemistry, answering quantitative questions that language models alone cannot address
- Callable from Claude and similar clients through an MCP server, so scientific simulations can be tried without building a dedicated compute cluster
- $2,000 in free credits for AI Simulation makes small-scale evaluation possible
- Broad coverage across drug discovery, materials, security, healthcare, and navigation from a single vendor
⚠️ Cons
- Most products are quote-only, making costs hard to estimate before engagement
- Aimed at enterprises and research institutions rather than everyday individual use
- Interpreting the output assumes domain knowledge in chemistry, materials, or cryptography
- The newest AQtive Guard capabilities are available to select customers first, with broader availability rolling out through later in 2026
- Japanese-language information and support are limited
Comparison with Similar Services
| Criterion | SandboxAQ | Schrödinger | Isomorphic Labs | Microsoft Azure Quantum Elements |
|---|---|---|---|---|
| Primary use | Drug discovery, materials, and security via LQMs | Physics-based molecular simulation | AI-driven drug discovery | Cloud platform for chemistry and materials research |
| Approach | Quantum technology + AI + scientific data | Physics computation + machine learning | Deep learning (AlphaFold lineage) | Integrated HPC + AI + quantum computing |
| Domains | Drug discovery, materials, healthcare, navigation, cryptography | Drug discovery, materials | Drug discovery | Chemistry, materials |
| Delivery | Product portfolio + MCP server + custom contracts | Software + joint research | Mainly partnerships with pharmaceutical companies | Cloud service on Azure |
| Price transparency | Partially published | Quote-driven | Undisclosed (partnership-based) | Azure usage-based pricing |
Who Is It For
- Researchers in drug discovery or materials development who want computational screening ahead of laboratory work
- Teams that want to screen catalysts or battery materials without maintaining their own compute infrastructure
- Anyone building conversational workflows that call scientific computation from an AI assistant such as Claude
- IT departments working on post-quantum cryptography migration or on gaining visibility and control over spreading AI use inside the organization
- Engineers who want to see what practical products are emerging at the intersection of quantum technology and AI
Summary
SandboxAQ approaches drug discovery, materials design, and security through “large quantitative models” that work with the numbers of physics and chemistry rather than with language. It is not a chat tool for casual individual use, but the release of its MCP server has made calling scientific simulations from Claude a far more realistic workflow than before. A practical starting point is to work within the free AI Simulation credits and check whether your own problem falls inside what LQMs cover.