ABACUS MCP Server

What is ABACUS MCP Server?

ABACUS MCP Server is a Model Context Protocol (MCP) server for ABACUS first‑principles calculations, designed for quantum chemistry and materials science. It provides an AI‑assisted interface to run SCF, structure optimization, molecular dynamics, band structure, density of states, and charge density calculations, with smart parameter suggestions and result analysis.

How to use ABACUS MCP Server?

Install Python 3.8+, ABACUS, and dependencies (fastmcp, ase, numpy, psutil). Configure the ABACUS_COMMAND, PSEUDO_DIR, and ABACUS_WORK_DIR environment variables or a config.json file. Start the server with python src/server.py and add it as an MCP server in clients like Claude Desktop or Roo Code. Use tools such as run_scf, create_structure, suggest_parameters, or diagnose_failure through natural‑language queries.

Key features of ABACUS MCP Server

• SCF, structure optimization, and molecular dynamics calculations
• Band structure, DOS, and charge density analysis
• Smart parameter suggestions and input validation
• Fault diagnosis and cost estimation
• Result interpretation with optimization recommendations
• PyABACUS integration for Python‑based workflows

Use cases of ABACUS MCP Server

• Running a SCF calculation on a silicon crystal with automatically suggested parameters
• Optimizing a perovskite geometry and computing its band gap
• Calculating the band structure of graphene along a specific k‑path
• Diagnosing convergence failures in SCF or MD simulations
• Using PyABACUS to analyze atomic orbital overlaps and Hamiltonian matrices

FAQ from ABACUS MCP Server

What are the system requirements?

Python 3.8+, the ABACUS software package, and sufficient computational resources. Pseudopotential files must be placed in a pseudos/ directory.

How do I configure ABACUS and the server?

Set environment variables ABACUS_COMMAND, PSEUDO_DIR, and ABACUS_WORK_DIR, or create a config.json file. Ensure ABACUS is on the system PATH and test with abacus --version.

What MCP clients are supported?

Claude Desktop and Roo Code are mentioned with example configurations. The server is started via the Python command python src/server.py.

What files are required for a calculation?

Pseudopotential files (e.g., Si.UPF) in the pseudos/ directory. The server can also fetch or generate appropriate files based on the calculation request.

How do I troubleshoot common problems?

Check which abacus, verify pseudopotential files exist in pseudos/, and for memory issues reduce ecutwfc or use a sparser k‑point grid. Set LOG_LEVEL=DEBUG for detailed logs.