Skip to main content
Glama

modelread

modelread_mcp_estimate_fit

Will this Hugging Face model run on this hardware, how fast, with which quant, and with what command? Checks the repo and its quantized derivatives (GGUF / MLX / AWQ / FP8) from their real file sizes and tensor headers. Returns the best variant and alternatives, each with: verdict (fits | fits-tight | offload | no | unknown), decode and prefill tokens/sec as a calibrated range (lo/mid/hi with basis), longest context that fits, memory placement (VRAM / RAM / SSD) and an exact llama.cpp / vLLM / MLX launch command.

Tool call template

This tool declares no required parameters. Add optional parameters from the input schema as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {},
    "name": "modelread_mcp_estimate_fit"
  }
}

Input Schema

NameRequiredDescriptionDefault
ctxNocontext length in tokens to fit (default 32768).
repoNoHugging Face repo id, e.g. "Qwen/Qwen3.6-27B" or "unsloth/gemma-4-26B-A4B-it-GGUF".Qwen/Qwen3.6-27B
hardwareNodevice preset id ("rtx4090", "m4max-128", "dgx-spark", "8x-h100-sxm", "iphone-17pro-12"),rtx4090
kv_cache_dtypeNoKV cache type: "f16", "q8_0" or "q4_0".f16

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observed

TDQS

A4.1/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description carries the full burden and does substantial work: it discloses the data source (real file sizes and tensor headers, implying an external metadata fetch), the full verdict taxonomy including "fits-tight", "offload" and "unknown", the uncertainty model (calibrated lo/mid/hi ranges with a stated basis), and what each result contains. It omits operational traits such as latency/cost of the check, caching, and any privacy/network implications of inspecting Hugging Face repos.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

It is front-loaded with the core question and every clause adds information (sources checked, verdict vocabulary, output fields, command families). The return-value enumeration packs many parenthetical items into one long sentence, which is dense but not padded or redundant.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

There is no output schema and no annotations, so the description must explain return values — and it does, enumerating verdict, speed ranges with basis, max context, VRAM/RAM/SSD placement, and the exact launch command. Minor gaps remain: the meaning of an "unknown" verdict, whether the call is purely read-only/metadata-only, and any runtime cost are not stated.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the four parameters are already well documented, and the description adds only partial meaning beyond that: it clarifies that "repo" may be a base repo or a GGUF-style quantized derivative, and that context length drives the fit verdict. It says nothing about how kv_cache_dtype or the hardware preset ids affect the estimate, leaving the baseline of 3 appropriate.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description opens with a concrete question ("Will this Hugging Face model run on this hardware, how fast, with which quant, and with what command?") and then states the mechanism (checking repo and GGUF/MLX/AWQ/FP8 derivatives via real file sizes and tensor headers). The single-repo × single-hardware-preset scope is orthogonal to the sibling set (search/find/list/compare), so an agent can discriminate it even though no sibling is named.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The framing question makes the usage context unmistakable: you have a specific repo id and a specific hardware preset and want a fit/speed/quant answer. There are no explicit exclusions or routing hints (e.g., nothing says to use modelread_mcp_compare_hardware when evaluating multiple devices), which keeps it below a 5.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources