ToxMCP Suite

ToxMCP is the public suite hub for guardrailed, auditable MCP servers across computational toxicology, exposure science, mechanistic reasoning, QSAR workflows, ADMET utilities, and downstream kinetic modeling.

If you want the single best place to follow the public suite, module launches, and onboarding path, star this repo.

• Preprint (bioRxiv): https://doi.org/10.64898/2026.02.06.703989
• Citation metadata: CITATION.cff

Why Star This Repo?

• It is the canonical public index for the ToxMCP suite.
• It is the best place to track which modules are public now versus still in progress.
• It is the suite-level home for docs, architecture, quickstart guidance, and contribution links.

Architecture

What Is MCP?

MCP (Model Context Protocol) is a standard way for LLM clients and orchestrators to call external tools over a structured interface. See the official MCP docs/spec: https://modelcontextprotocol.io/

Public Suite Today

Getting Started

The fastest way to get productive with ToxMCP is:

1. Pick an entry point. Start with CompTox MCP for the most recognizable public toxicology data workflow, Direct-Use Exposure MCP for product-use exposure scenarios, or Dietary Exposure MCP for food-mediated oral screening.
2. Follow that module repo’s Quickstart TL;DR.
3. Run its Verification (smoke test) so you know the local setup is working before adding more modules.

Common Local Ports

• CompTox MCP: http://localhost:8000
• O-QT MCP: http://localhost:8001
• PBPK MCP: http://localhost:8002
• AOP MCP: http://localhost:8003
• ADMETlab MCP: http://localhost:8200

Which one should I try first?

• Choose CompTox MCP if you want the best first-run experience and the most recognizable toxicology dataset integration today.
• Choose Direct-Use Exposure MCP if you need reviewable exposure scenarios with explicit assumptions, provenance, limitations, and fit-for-purpose framing.
• Choose Dietary Exposure MCP if you need screening-only food-mediated oral exposure calculations, governed residue and consumption review, or a PBPK-ready oral dose handoff.
• Choose Environmental Fate MCP if you need auditable environmental release-to-concentration screening, scalar erosion/sediment transport screening, scientific review packets, or downstream concentration handoff packages.
• Add AOP MCP, O-QT MCP, or PBPK MCP next depending on whether the workflow needs mechanistic context, QSAR Toolbox automation, or toxicokinetic modeling.

The direct-use, dietary, and environmental-fate exposure layers are now public. Direct-Use Exposure MCP covers deterministic product-use external-dose screening; Dietary Exposure MCP v0.1.0 covers food-mediated oral screening and governed evidence handoffs; and Environmental Fate MCP covers bounded environmental release-to-concentration screening. Each remains a bounded screening service rather than a final safety or regulatory decision engine.

Docs

Start with docs/README.md for the suite-level documentation map.

Notes

• Each MCP server is versioned and released independently.
• Some modules depend on proprietary or rate-limited upstream services. Check each module README for exact setup details.

Acknowledgements / Origins

ToxMCP was developed, in part, in the context of the VHP4Safety project (see: https://github.com/VHP4Safety) and related research and engineering efforts.

Funding: Dutch Research Council (NWO) — NWA.1292.19.272 (NWA programme)

This suite integrates with third-party data sources and services (e.g., EPA CompTox, ADMETlab, AOP resources, OECD QSAR Toolbox, Open Systems Pharmacology). Those upstream resources are owned and governed by their respective providers; users are responsible for meeting any access, API key, rate limit, and license/EULA requirements described in each module.

License

Apache-2.0

✅ Citation

Djidrovski, I. ToxMCP: Guardrailed, Auditable Agentic Workflows for Computational Toxicology via the Model Context Protocol. bioRxiv (2026). https://doi.org/10.64898/2026.02.06.703989

@article{djidrovski2026toxmcp,
  title   = {ToxMCP: Guardrailed, Auditable Agentic Workflows for Computational Toxicology via the Model Context Protocol},
  author  = {Djidrovski, Ivo},
  journal = {bioRxiv},
  year    = {2026},
  doi     = {10.64898/2026.02.06.703989},
  url     = {https://doi.org/10.64898/2026.02.06.703989}
}

Citation metadata: CITATION.cff

Contributing

Please see CONTRIBUTING.md. If you’re not sure which repository to use, open an issue here: https://github.com/ToxMCP/toxmcp/issues

Support

See SUPPORT.md.