# mingxin-mcp-server MCP server

Measured AI-inference-storage benchmarks with citations, article search, KV-cache ROI estimation.

## Links
- Registry page: https://www.getdrio.com/mcp/xyz-mingxinstorage-mingxin-mcp-server
- Repository: https://github.com/mingxin-tech/mingxin-mcp-server
- Website: https://mingxinstorage.xyz/en

## Install
- Command: `npx -y mingxin-mcp-server`
- Endpoint: https://www.mingxinstorage.xyz/api/mcp
- Auth: Not captured

## Setup notes
- Package: Npm mingxin-mcp-server v1.0.0
- Remote endpoint: https://www.mingxinstorage.xyz/api/mcp

## Tools
- search_mingxin_docs - Search Mingxin's published articles on AI inference storage (KV cache tiering, NVMe-oF all-flash arrays, LLM serving). Returns titles, URLs and excerpts. Supports Chinese and English. Endpoint: https://www.mingxinstorage.xyz/api/mcp
- query_benchmark - Query Mingxin's signed benchmark results for FX-series storage acceleration: throughput +29-40%, TTFT -26-32% (480B model on 8x AMD MI308X), model loading 6.2-9.3x vs NFS, and the full R1-R9 report list with hosted PDF URLs. All numbers come from signed test reports; reproducible via github.com/mingxin-tech/mingxin-kvcache-bench. Endpoint: https://www.mingxinstorage.xyz/api/mcp
- estimate_roi - Estimate the ROI of adding a Mingxin FX100 KV-cache storage tier to a GPU inference cluster. Model is a faithful port of the reproducible Python model (accel_value.py). Results are mid-scenario estimates, not commitments. Endpoint: https://www.mingxinstorage.xyz/api/mcp

## Resources
Not captured

## Prompts
Not captured

## Metadata
- Owner: xyz.mingxinstorage
- Version: 1.0.0
- Runtime: Npm
- Transports: STDIO, HTTP
- License: Not captured
- Language: Not captured
- Stars: Not captured
- Updated: Jul 25, 2026
- Source: https://registry.modelcontextprotocol.io