# fitllm.run — software development services — 2 of 5

Home > Browse > By type > tool > fitllm.run

> Probe-verified (tier 2 of 3). Last checked 2026-09-26. 2 of 5 capability checks passed. 2 unknown.

- host: fitllm.run
- url: https://fitllm.run/
- category: software development services
- type: tool
- language: en

## Probe-verified · tier 2 of 3

Probed live on 2026-09-26 against rubric v1.2. 2 of 5 checks passed. 2 unknown.

- UNKNOWN — Markdown negotiation (`Accept: text/markdown`)
- UNKNOWN — Markdown twin at a predictable URL (`/page.md`)
- PASS — llms.txt (`/llms.txt`)
- FAIL — API catalog (`RFC 9727`)
- PASS — MCP endpoint answers tools/list (`initialize → tools/list, no credentials`) · on 26 Sep 2026
  - ABSENT — MCP server card (`/.well-known/mcp/server-card.json`): absent; not scored
  - Platform-supplied: no; not scored
- FAIL — Other agent files at well-known paths (`agent-skills/index.json · oauth-authorization-server · oauth-protected-resource · http-message-signatures-directory`): none found; not scored

MCP tools (from tools/list on 2026-09-26): check_llm_fit, what_fits_on_hardware, list_supported

## Summary

Auditable calculator estimating if local LLMs fit on NVIDIA GPUs or Apple Silicon Macs by modeling memory use including weights, KV cache, and runtime overhead for architectures like Gemma 4, Qwen 3.6, and GLM families.

## Context

2370 sites in this index share the category "software development services"; 246 are transactional.

## How we grade

- tier 1 — **Declared**: The site says agents are welcome in its robots.txt. A permission, not a capability.
- tier 2 — **Probe-verified** (this site): We fetched the site and tested each signal ourselves. Dated, repeatable, re-checked on a schedule.
- tier 3 — **Behaviourally verified**: An agent completed the action end to end — booked, bought, submitted — and it worked.

## About this record

Known Good probes sites live and publishes what was tested and when.
Method: https://knowngood.sh/about.md · Rubric: https://knowngood.sh/benchmark

How we found this site: probe discovery. This site didn't declare agent access — we found it by checking the open web. Everything above was tested the same way, against the same rubric, on the date shown. [How we choose what to check, and how to opt out](https://knowngood.sh/bot)

## Where this sits

- tool: https://knowngood.sh/type/tool
- software development services: https://knowngood.sh/category/software-development-services
- Browse: https://knowngood.sh/browse
- Rubric: https://knowngood.sh/benchmark