MCP.so
登录

FitLLM

@click6067-ship-it

关于 FitLLM

Will this LLM fit on your GPU, multi-GPU rig or Mac? Exact VRAM & KV-cache math. Remote server: https://fitllm.run/api/mcp

基本信息

分类

开发工具

许可证

MIT

运行时

node

发布者

click6067-ship-it

提交者

Yo

配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "fitllm-engine": {
      "command": "npx",
      "args": [
        "fitllm",
        "GLM-4.7-Flash",
        "--gpu",
        "4090",
        "#",
        "✓",
        "FITS",
        "—",
        "21.9/24",
        "GB,",
        "free",
        "2.1",
        "GB"
      ]
    }
  }
}

工具

未检测到工具

工具是从 README 中自动提取的。维护者可以在 ## Tools 标题下列出工具,即可填充这部分内容。

概览

What is Fitllm Engine?

Will this LLM fit your GPU or Mac? npx fitllm — accurate memory math for MLA/sliding-window/hybrid/MoE architectures that naive VRAM calculators get wrong by up to 18x. Single file, zero deps, conformance-vector tested. MIT.

How to use Fitllm Engine?

The README includes setup instructions such as npx fitllm "GLM-4.7-Flash" --gpu 4090 # ✓ FITS — 21.9/24 GB, free 2.1 GB.

Key features of Fitllm Engine

  • Architecture values checked against official HuggingFace config.json
  • Calibration: Qwen 3.6 35B-A3B @128K, 8-bit ≈ 54 GB (matches real local runs)
  • MLA per-token cost: GLM-4.7-Flash = (512 + 64) × 2 B × 47 layers = 54,144 B/token — pinned by conformance vectors
  • Claude Code: claude mcp add --transport http fitllm https://fitllm.run/api/mcp
  • Cursor / Windsurf: add to mcp.json → { "mcpServers": { "fitllm": { "url": "https://fitllm.run/api/mcp" } } }

Use cases of Fitllm Engine

  • Connect an MCP-compatible client to this repository's service.
  • Review the README-backed setup before enabling it in production.

FAQ from Fitllm Engine

Where is the source code for Fitllm Engine?

The source code is linked from the repository URL on this page.

Does Fitllm Engine include a standard MCP config?

If the README contains a parseable MCP configuration block, it is shown in the Config tab.

评论

开发工具 分类下的更多 MCP 服务器