Read-only: which local LLMs a GPU or Mac can run - VRAM fit, tokens/sec, model specs.
Scored across 5 tools
Each tool targets a distinct query type: id lookup, single model specs, model-machine fit, models for a machine, and machine comparison. The descriptions explicitly delimit boundaries, e.g. check_hardware_fit vs models_for_hardware, with no real overlap.
Four of five tools follow a clear verb_noun pattern (check_hardware_fit, compare_hardware, get_model_specs, search_catalog). models_for_hardware breaks that pattern with a noun-first name, but it is readable and not confusing.
Five tools are well-scoped for a read-only advisory server about local LLM hardware and models. Each tool earns its place: one resolver plus four distinct analytical queries, with no redundancy or bloat.
Core workflows (search, specs, fit, machine comparison, models for a machine) are covered. However, notable gaps exist: no tool to get single-hardware specs without comparison, no direct hardware-for-model query, and no model-to-model comparison, which are natural questions in this domain.
Use this when the user asks whether one specific model runs on one specific machine, at what memory cost, and roughly how fast. Needs a hardware id and a model id from search_catalog. Returns a fit verdict, the memory arithmetic, and a decode-speed range with a confidence label. Memory is sized from all parameters, speed from active parameters. It does not recommend purchases and does not run anything.
Inputs: model_id (required), hardware_id (required), quantisation, context_length. Returns: url, model, notes, speed, +8 more
Use this when the user compares two to four machines. Needs hardware ids from search_catalog. Returns memory, bandwidth, memory type, release year, status and a dated price where one is recorded, side by side. A price of null means none is on record. It does not check any model against the machines; use check_hardware_fit for that.
Inputs: hardware_ids (required). Returns: summary, hardware, data_as_of
Use this when the user asks about one model's size, architecture, context window or licence. Needs a model id from search_catalog. Returns total and active parameters, weight size per quantisation, context length, licence, release date and sources. It does not say whether the model fits any machine; use check_hardware_fit for that.
Inputs: model_id (required). Returns: url, model, sources, summary, +2 more
Use this when the user asks what models a given machine can run. Needs a hardware id from search_catalog. Returns up to 20 catalogue models ranked for that machine, each with a fit verdict and a speed range, optionally filtered to one use such as coding. It covers only models in the catalogue, at one quantisation and context length per call, and leaves out models that do not fit.
Inputs: hardware_id (required), limit, use_case, quantisation, +1 more. Returns: url, models, summary, hardware, +4 more
Use this first when the user names a GPU, Mac, mini PC or language model and you need its id for the other tools. Matches names and common aliases in a fixed catalogue of local-LLM hardware and models. It does not return specs, fit or speed, and it does not search the web.
Inputs: kind (required), query (required), limit. Returns: kind, matches, summary, data_as_of
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Will this LLM fit on your GPU, multi-GPU rig or Mac? Exact VRAM & KV-cache math. Read-only.
Measured local-LLM tok/s on home AI rigs. Search builds, can-it-run, best GPU; sourced + linked.
Can this LLM run on my GPU? VRAM, speed ceiling and what fits instead, for any model and GPU.
Read-only GPU advisor: model VRAM needs, fit checks and rental prices, with evidence grades.
Add one secure layer between your agents and this server.
