Changelog

All notable changes to this project will be documented here.

The format follows Keep a Changelog.

[v0.1.0] - 2026

Added

• Core MCP server with stdio transport
• list_gpus tool — enumerate all GPUs with utilization and memory
• get_gpu_metrics tool — detailed per-GPU metrics (temp, power, PCIe, NVLink)
• get_gpu_processes tool — PID-level GPU process attribution
• gpu_summary tool — aggregate stats across all devices
• NVML-based collector with MIG support
• Mock collector for testing without GPU hardware
• --version flag
• GitHub Actions CI (build, test, lint)
• Helm chart workflow
• Dockerfile (multi-stage build)
• Helm chart under deploy/helm/gpu-mcp-server
• server.json MCP registry manifest
• Apache 2.0 license
• GOVERNANCE.md (LF Minimum Viable Governance)
• SECURITY.md (private disclosure, 48-hour response)
• AI_GUIDELINES.md (AI contribution policy)
• AGENTS.md (instructions for AI agents)