Materials MCP Commons
Provides an independently deployable Cloudflare Worker boundary that hosts the canonical schema hostname, serving the project's versioned JSON Schema contracts.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Materials MCP Commonsfind the crystal structure of silicon in OPTIMADE"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Materials MCP Commons
Materials MCP Commons is an open, interoperable foundation for exposing materials-science data, simulations, workflows, and computing capabilities through the Model Context Protocol (MCP).
The project is being built around six invariants:
a plugin-agnostic engine before any provider- or solver-specific plugin;
real, provenance-traceable scientific data and real scientific systems;
versioned, language-neutral contracts before backend-specific adapters;
focused federated servers rather than one unrestricted server;
bounded model context through progressive capability discovery and artifact references;
evidence-backed Conformant, Verified, and Validated trust levels.
Status
Active development. No production-alpha release has been published yet.
The first release is intended to be a research-grade production alpha with a deliberately bounded, evidence-complete capability set. Alpha denotes API and scope evolution; it does not relax scientific-validity, provenance, security, or reproducibility requirements.
Related MCP server: production-grade-mcp-agentic-system
Engine-first scope
Materials MCP Profile and conformance tooling
Generic plugin manifest, lifecycle, dispatch boundary, and Scientist Kit
Context Gateway minimum
Shared artifact, structure, and durable-run services
Generic effect, permission, plan, provenance, and policy enforcement
Production packaging, observability, recovery, and host interoperability
The engine was implemented and gated before the first concrete integration was selected. Concrete plugins are separately packaged and must not become dependencies or special cases in the engine core. The active first proof is the separate full OPTIMADE integration under plugins/optimade/; it uses only real providers and provenance-traceable data. Mocked scientific backends and fabricated scientific results are not accepted as implementation or release evidence.
Documentation
Public documentation begins in docs/.
The repository now contains:
an installable, typed Python alpha package with audited generic JSON Schema runtime dependencies;
the immutable 0.1.0 contract line and additive 0.2.0 successor, each with a checksum-pinned resource index;
an offline, fail-closed conformance harness;
a deterministic multi-version conformance report and structural zero-coupling audit;
strict offline profile/package validation and deterministic registration, discovery, inspection, and activation leases;
exact-owner dispatch with immutable typed requests, two-sided schema validation, and contract-valid structured failures;
deterministic operation plans, exact scoped permissions, expiring single-use approvals, atomic quotas, atomic at-most-once receipt redemption, and tamper-evident policy audit;
canonical context measurement with identified host tokenizers, bounded linting, real-source compact projection, retrieval evaluation, and long-session lease checks;
deterministic empty-workspace scaffolding and exact-profile declarative package building without generated capabilities, handlers, or scientific results;
deterministic plugin-package and versioned compatibility reports, explicit migration assessment, and manifest-derived capability references without plugin-code execution;
durable owner-isolated run events, reconnect/cancellation, and hash-verified artifact/provenance records;
an optional official-SDK MCP adapter with four bounded tools, trusted host-side identity/authorization injection, and real in-process/stdio protocol tests;
metadata-only operational events, bounded health metrics, and integrity-checked non-overwriting SQLite state snapshots/restores;
one registered real positive scientific record, its lossless successor migration and bounded compact projection, and an isolated generated-negative rejection corpus; and
an independently deployable Cloudflare Worker boundary for the canonical schema hostname; and
the separately packaged, exact-profile OPTIMADE integration contract, with seven bounded capabilities and no provider coupling in the engine.
Start with the runtime guide, production-alpha API compatibility guide, support boundary, MCP host guide, operations runbook, state recovery guide, threat model, Scientist Kit authoring guide, contract guide, versioning policy, scientific evidence policy, and schema-hosting guide. Effectful dispatch is available only through the exact policy boundary; remote scheduling and network-host authentication are not implemented. The OPTIMADE integration's checked support matrix separates local observations from immutable-commit hosted evidence. The hosted matrix and manual live-provider run pass for runtime commit 071b341da1766388689be92f74a894444970e493; no package release or Validated claim is made.
Author and maintainer
Hesham Salama
Licensing
Source code is licensed under the Apache License 2.0. Project-authored documentation and specifications are licensed under CC BY 4.0. Third-party dependencies, datasets, models, and scientific systems retain their own terms.
This server cannot be deployed
Maintenance
Related MCP Connectors
Hosted MCP server for AI agent identity, permissions, verification, and reusable proof.
MCP server for mandates, delegation, policy-gated execution, credential grants, and audit.
Capability registry for the agentic economy. Semantic search over verified MCP server listings.
- UnifAPIOAuthcom.unifapi
Hosted MCP server for live public-data APIs and Skills for AI agents.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceA lightweight, fast MCP server that provides onchain capabilities for the LLMs and Agents.39 npm253MIT
- AlicenseNot gradedqualityDmaintenanceA production-grade MCP server designed for multi-tenant, authenticated, and observable AI agent systems, enabling secure tool execution across heterogeneous data sources.62MIT

Local AI MCPofficial
AlicenseAqualityAmaintenanceUnified MCP server for managing local model runtimes (Ollama, LM Studio, etc.), enabling provider-agnostic discovery, lifecycle management, hardware-fit checks, and delegated inference.1618 npmCreative Commons Attribution Non Commercial No Derivatives 4.0 International- AlicenseNot gradedqualityBmaintenanceUnified control plane and runtime for the Model Context Protocol ecosystem, enabling plugin management, configuration, permissions, and API gateway services.20 npm1MIT