Colabfit Mcp
Connects Claude to the ColabFit materials science database and lets you train MACE interatomic potentials locally through conversational commands. You get tools to search and download atomic structure datasets from HuggingFace, kick off GPU or CPU training jobs via KLIFF, run energy and force calculations on the trained models, and validate them against OpenKIM test drivers for properties like elastic constants and vacancy formation energies. The whole thing runs in Docker with automatic GPU passthrough when available. Works over stdio with any MCP client. Useful when you're prototyping machine learning potentials for materials and want to skip writing pipeline scripts, or when you need to quickly pull training data and benchmark a model against standard tests without leaving your AI chat.
- GitHub stars
- 2
- tools
- 9
- transport
- STDIO
- last updated
- Oct 16, 2025
Tools
9 toolsPublic tool metadata for what this MCP can expose to an agent.
search_datasetsSearch ColabFit database by text, elements, properties, software
Search ColabFit database by text, elements, properties, software
No parameter schema in public metadata yet.
check_local_datasetsScan local data directory for downloaded datasets, filter by elements/properties
Scan local data directory for downloaded datasets, filter by elements/properties
No parameter schema in public metadata yet.
download_datasetDownload a dataset from HuggingFace via KLIFF
Download a dataset from HuggingFace via KLIFF
No parameter schema in public metadata yet.
train_maceTrain a MACE-style KLAY model from scratch using KLIFF
Train a MACE-style KLAY model from scratch using KLIFF
No parameter schema in public metadata yet.
use_modelRun energy/forces/relax calculations with a trained KLAY model, or generate a Python snippet
Run energy/forces/relax calculations with a trained KLAY model, or generate a Python snippet
No parameter schema in public metadata yet.
check_statusCheck GPU, packages, disk, existing models and datasets
Check GPU, packages, disk, existing models and datasets
No parameter schema in public metadata yet.
list_test_driversList available kimvv test drivers, optionally filtered by property keyword
List available kimvv test drivers, optionally filtered by property keyword
No parameter schema in public metadata yet.
run_test_driverRun a kimvv test driver against a trained KLAY model; saves structures.extxyz + results.json in a timestamped subdirectory; supports multiple structures per call with optional repeat for supercell sizing and async_mode for slow drivers
Run a kimvv test driver against a trained KLAY model; saves structures.extxyz + results.json in a timestamped subdirectory; supports multiple structures per call with optional repeat for supercell sizing and async_mode for slow drivers
No parameter schema in public metadata yet.
check_test_driver_resultCheck status of an async test driver job and return inline results when complete
Check status of an async test driver job and return inline results when complete
No parameter schema in public metadata yet.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageReadme
colabfit-mcp
An MCP server for discovering ColabFit datasets and training MACE interatomic potentials using KLIFF and KLAY.
Overview
This is a Model Context Protocol (MCP) server that gives AI assistants the ability to:
- Search and download scientific datasets from ColabFit
- Train MACE interatomic potentials on your local hardware (GPU or CPU)
- Run energy/forces calculations and validate models with OpenKIM test drivers
It bridges conversational AI and local compute — the AI agent searches for data, trains models, and runs simulations on your machine through this server.
Prerequisites
- Docker and Docker Compose v2 — for the containerized server
- Git — for cloning the repository
- make — for the quick-start commands (optional; manual steps are documented below)
- (Optional) NVIDIA GPU + drivers — for GPU-accelerated training
- (Optional) nvidia-container-toolkit — required for Docker to access the GPU
Details
- Repository
- github.com/colabfit/colabfit-mcp
- Category
- Databases
- Package
colabfit-mcp- Transport
- STDIO
- Registry
- MCP registry, active
Related Databases MCP servers
View all
Delx Memorydavidmosiah/delx-memoryLocal-first persistent memory MCP: shared SQLite key/value store, searchable, TTL-aware, secret-safe2stars
Lorebaseaantenore/lorebaseRead and change a governed, versioned knowledge graph on Postgres through typed tools.2stars
Sentinal Mcpmusaraf-m/sentinalRedis health, BullMQ queue monitoring, memory analysis, and slow query diagnostics2stars
Mcp Server Spreadsheetmarekrost/mcp-server-spreadsheetMCP server for spreadsheet files (.xlsx, .csv, .ods) - cell-level ops + DuckDB SQL engine.2stars
Snowflake MCP Servernsphung/mcp-snowflake-serverA Snowflake MCP server — SQL queries, schema exploration, and data insights for AI assistants2stars
LogBearteralabx/logbearLocal dev server logs for coding agents: errors, runs and diffs as read-only SQL.2stars