MCP tools for autonomous materials research
Lattice Graph exposes its full platform capability as MCP tools, enabling AI agents to conduct materials research autonomously. This recorded session walks through the tool catalog, integration patterns, and real workflow examples.
What are MCP tools?
The Model Context Protocol (MCP) provides a standardized interface for AI models to invoke external tools. Lattice Graph publishes MCP tools that cover materials search, property lookup, confidence scoring, simulation dispatch, patent analysis, and report generation. Any MCP-compatible agent (Claude, custom LangGraph agents, or third-party frameworks) can use these tools.
Core search and lookup tools
The foundational tools include materials_search (formula and property queries), get_material_properties (detailed property retrieval with provenance), confidence_score (cross-source agreement for a material-property pair), and similar_materials (find structurally or compositionally similar candidates). These tools return structured data with source attribution attached.
Simulation and analysis tools
Advanced tools enable agents to run formation energy predictions, dispatch PyBaMM battery simulations, generate CTE stack analyses, and request volcano plot data for catalyst screening. Each tool validates inputs and returns results with uncertainty estimates and provenance chains.
Building autonomous workflows
The real power of MCP tools emerges when agents chain them into multi-step workflows: search for candidates, score confidence, check patent coverage, evaluate supply chain risk, run simulations, and compile results into a report. Lattice Graph's LangGraph-based agent system orchestrates 12 specialist agents that use these tools to handle complex research questions end to end.
Ready to try it yourself?
Start exploring materials data with cross-source confidence, simulation workflows, and the tools described in this resource.