claude-perf-tune

cfregly/claude-perf-tune

★ 0 stars Python AI/LLM Updated today

31 GPU inference profiling and optimization skills for Claude Code, with a bundled MCP server

View on GitHub → Try with Claude — $10 free →

Quick Install

Copy the config for your editor. Some servers may need additional setup — check the README.

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "claude-perf-tune": {
      "command": "uvx",
      "args": [
        "claude-perf-tune"
      ]
    }
  }
}

Or install with pip: pip install claude-perf-tune

README Excerpt

GPU inference profiling and optimization skills for [Claude Code](https://claude.com/claude-code), backed by a bundled MCP server: shipped as the `profile-and-optimize` plugin. 31 task-oriented workflows covering benchmark sweeps, kernel-level profiling (nsys / ncu / DCGM / zymtrace), speed-of-light roofline analysis, quantization and speculative-decode tuning, and a multi-page PDF perf-tune report renderer. Each skill is a `SKILL.md` following the open [Agent Skills standard](https://agentskill

Topics

agent-skillsclaude-codecudagpuinferencellmmcpperformance

claude-perf-tune

Quick Install

README Excerpt

Topics

Related AI/LLM Servers