# Run It Local

> A free, sourced guide to running AI on hardware you control: daily local-AI news, checked prices on local-AI hardware (GPUs, Macs, AI mini PCs, RAM, storage), and a private toolkit for deciding which AI jobs stay in-house. Published by FolkTech AI. Nothing on this site is for sale. This is a private preview: not launched and not indexed.

The site informs; it does not sell. No accounts, no cookies, no analytics. All data below is free to read and query, read-only.

## Data freshness

- Hardware deals: refreshed every morning around 6 AM PT. Last refresh: Oct 18, 2026 8:55 AM PT (422 tracked prices: GPUs 29, Apple (Mac Studio, Mac mini, Mac Pro, MacBook Pro) 61, Mini PCs, AI boxes and workstations 121, RAM 88, SSDs and NAS storage 123). Big-retailer searches rotate, so each is re-checked at least every 3 days; every item carries its own last_seen time.
- News: refreshed every morning around 6 AM PT. Last refresh: Oct 18, 2026 6:02 AM PT (26 stories, last 45 days).
- Guides, toolkit, and protocols: edited by hand; see the modified date in /api/guides.json.

## Editorial rules

- Tracked deals are chosen and ordered only by price data we check ourselves (price, then specs). Money never changes which items are listed or how they rank.
- Anything a company pays for is labeled Sponsored and kept in a separate slot, outside the tracked list, the filters, this data, and the MCP tools.
- If any product links become affiliate links later, they will be labeled, and affiliate status will never change rankings or editorial picks.
- Scope is strictly hardware for running AI locally: GPUs, AI mini PCs/boxes and workstations, Mac Studio / Mac mini / Mac Pro / MacBook Pro, RAM, SSDs and NAS. No monitors, iMacs, MacBook Air, accessories, or cloud services.
- Prices come only from the retailer's or maker's own page or official feed (Amazon, Best Buy, Newegg, B&H, Micro Center, Apple, and official maker stores). Never marketplaces or private sellers.
- A deal flag means the price is below our own tracked baseline or the retailer itself shows a higher original price. Codes appear only if printed on the retailer's page.
- News headlines link to the original source; summaries are our own one-line descriptions, not quotes. Inclusion is not an endorsement.

## Interact (MCP server, read-only)

- [MCP endpoint](https://runitlocal.app/mcp): Streamable HTTP, JSON responses, no auth, no accounts. Supports protocol 2026-07-28 (stateless, server/discover) and the older initialize handshake (2025-11-25, 2025-06-18, 2025-03-26).
  Tools: search_deals (category, max_price, min_price, min_ram_gb, min_vram_gb, min_storage_gb, min_bandwidth_gbps, retailer, apple_line, query, deals_only, sort [price_asc|price_desc|memory_desc|bandwidth], limit), get_chip_specs (query), get_news (limit, query, since), list_guides (section, query).
  Toolkit tools (stateless: computed from your inputs, nothing stored or logged; same scoring and math as the site's pages):
  place_job (sensitivity, complexity, frequency, outage, ownership: 0-2 each) -> Local / Hybrid / Cloud with signal scores;
  readiness_check (no args -> the 21 statements; answers: 21 x No|Partly|Mostly|Yes|Not sure, optional previous) -> score 0-100, band, areas, first protocols;
  cost_scenario (users, seat, api, hardware, license, support, years) -> cloud vs local totals and indicative hardware recovery;
  capacity_scenario (people, minutes, occurrences, rate) -> modeled hours and value per month/year;
  get_protocol (id P-00..P-08; omit to list) -> the full checklist.
- [Server card](https://runitlocal.app/mcp/server-card) and [AI catalog](https://runitlocal.app/.well-known/ai-catalog.json) (MCP Server Card discovery, SEP-2127).
- [Plain HTTP search](https://runitlocal.app/api/search?category=box,apple&min_ram_gb=64&min_storage_gb=1000&max_price=2500&sort=bandwidth): same filters as search_deals, as GET query parameters, returns JSON.

## Open data (static JSON, refreshed daily)

- [Deals](https://runitlocal.app/api/deals.json): every tracked price with category, line, retailer, price, specs, url, last_seen, deal_flag. Schema: [deals.schema.json](https://runitlocal.app/api/schema/deals.schema.json)
- [News](https://runitlocal.app/api/news.json): title, url, source, published, summary. Schema: [news.schema.json](https://runitlocal.app/api/schema/news.schema.json)
- [Guides](https://runitlocal.app/api/guides.json): every page with title, description, section, url.
- [Protocols](https://runitlocal.app/api/protocols.json): the nine P-00 to P-08 checklists as structured JSON (indications, equipment, steps, checks, handoff, sources).
- [Toolkit definitions](https://runitlocal.app/api/toolkit.json): the questions, options, example inputs and disclaimers behind the toolkit tools.
- [Chip specs](https://runitlocal.app/api/chips.json): our hand-maintained table of memory bandwidth (GB/s) per chip or GPU. Every figure is the maker's published spec with its source URL, not a benchmark we ran; unknowns are null.
- [OpenAPI description](https://runitlocal.app/api/openapi.json) of the JSON files and the search endpoint.

## Spec fields on each deal

- specs.storage_gb: system SSD for mini PCs, Macs and workstations; drive capacity for SSDs (capacity_gb also kept). Taken from the listing title, the retailer's product page, or the retailer URL. null when the listing doesn't state it; we never guess.
- specs.chip and specs.memory_bandwidth_gbps: from /api/chips.json. Set only for unified-memory systems (Apple silicon, AMD Ryzen AI Max+ 395 / PRO 495, NVIDIA GB10) and discrete GPU cards. Mini PCs with socketed or vendor-chosen DDR5 stay null because their bandwidth depends on the installed memory.

## Sections

- [Home](https://runitlocal.app/): what the site is, latest news, top tracked deals, and the toolkit entry points.
- [Hardware deals](https://runitlocal.app/deals.html): checked prices on GPUs; Apple (Mac Studio, Mac mini, Mac Pro, MacBook Pro, one collapsible group per line); mini PCs, AI boxes and workstations; RAM (64 GB+ DDR5); SSDs (1 TB+) and NAS. Sponsored slot is separate and labeled. All listings are in the HTML; the filter bar is optional JavaScript. Deep links: deals.html?cat=box&mem=64&storage=1000&max=2000&ret=Newegg
- [AI news](https://runitlocal.app/news.html): recent stories and discussions about local, on-prem, and private AI, linked to the original source with our own one-line summary.
- Toolkit for teams ("Does your AI need the cloud?"): runs entirely in the browser; nothing entered is sent anywhere.
  - [Capacity scenario](https://runitlocal.app/capacity.html): Make the cloud a choice. A free, private toolkit for deciding which AI jobs stay in-house, getting the knowledge behind them ready, and modeling the tradeoff.
  - [Readiness check](https://runitlocal.app/check.html): A free, private, 10-minute check of how ready your company knowledge is for AI on hardware you control.
  - [Cost scenario](https://runitlocal.app/cost.html): Make the cloud a choice. A free, private toolkit for deciding which AI jobs stay in-house, getting the knowledge behind them ready, and modeling the tradeoff.
  - [Model it](https://runitlocal.app/model.html): Make the cloud a choice. A free, private toolkit for deciding which AI jobs stay in-house, getting the knowledge behind them ready, and modeling the tradeoff.
  - [Place it](https://runitlocal.app/place.html): Mike Folk's workload placement guide: score one AI job on data, task, frequency, reliability, and ownership.
  - [Protocols](https://runitlocal.app/protocols.html): Make the cloud a choice. A free, private toolkit for deciding which AI jobs stay in-house, getting the knowledge behind them ready, and modeling the tradeoff.
- [Protocols](https://runitlocal.app/protocols.html): short step-by-step fixes for getting a team's knowledge ready for in-house AI.
  - [P-00 The 20-question drill](https://runitlocal.app/protocols/p00-twenty-question-drill.html): Test whether your documents can answer the questions people actually ask. Run it before you buy anything.
  - [P-01 Establish one source of truth](https://runitlocal.app/protocols/p01-one-source-of-truth.html): Get to one current, approved copy of the handbook and core policies, in one place.
  - [P-02 Assign owners and review dates](https://runitlocal.app/protocols/p02-owners-and-review-dates.html): Give every important document a person responsible for it and a date it gets looked at again.
  - [P-03 Map who sees what](https://runitlocal.app/protocols/p03-map-who-sees-what.html): Sort your knowledge into everyone, managers, and restricted, and make permissions match.
  - [P-04 Capture tribal knowledge once](https://runitlocal.app/protocols/p04-capture-tribal-knowledge.html): Record the people who know how things work explaining it once, then turn it into written steps.
  - [P-05 Draw the red line](https://runitlocal.app/protocols/p05-draw-the-red-line.html): Write a one-page AI use policy: what stays in-house, what AI may do, and what it won't.
  - [P-06 Run a desk pilot](https://runitlocal.app/protocols/p06-desk-pilot.html): Try handbook Q&A on one machine with open-source tools before buying anything.
  - [P-07 Size the closet tier](https://runitlocal.app/protocols/p07-size-the-closet.html): Move from one desk to one dedicated machine on your network, sized from what the pilot showed.
  - [P-08 Redact before the cloud](https://runitlocal.app/protocols/p08-redact-before-cloud.html): When a cloud tool is the right choice, strip personal details on your own machine first.

## About

- [About](https://runitlocal.app/about.html): Make the cloud a choice. A free, private toolkit for deciding which AI jobs stay in-house, getting the knowledge behind them ready, and modeling the tradeoff.
- [Sponsor](https://runitlocal.app/sponsor.html): Sponsor the weekly protocol. Partners in local AI hardware, models, and tools are welcome.
- [Who’s behind this](https://runitlocal.app/who.html): Who writes this site, why, and the editorial standards it follows: sourced, vendor-neutral, disclosed, and private by design.
- [Write for us](https://runitlocal.app/write-for-us.html): Make the cloud a choice. A free, private toolkit for deciding which AI jobs stay in-house, getting the knowledge behind them ready, and modeling the tradeoff.
- [Protocol P-00: The 20-question drill](https://runitlocal.app/issues/001.html): Make the cloud a choice. A free, private toolkit for deciding which AI jobs stay in-house, getting the knowledge behind them ready, and modeling the tradeoff.

## Optional

- [llms-full.txt](https://runitlocal.app/llms-full.txt): this file plus the full current deals and news lists in plain text.
- [sitemap.xml](https://runitlocal.app/sitemap.xml)
Contact: feedback@folktechai.com