For agents

Ask your AI assistant which box to buy, and it answers from our latest prices and published speeds. Free, no key, no account. Prices last checked Oct 24, 12:11 UTC. We check once a day.

Questions your agent can answer

Ask	Tool
What's the cheapest machine that runs gpt-oss-120b?	`find_for_model`
Show me 128GB boxes under $3,000, fastest first.	`search_machines`
Is the Framework Desktop cheaper on Amazon or direct?	`get_machine`
Did anything drop in price since Monday?	`list_changes`
Which models do you have data for?	`list_models`

Answers link back to our machine pages and include when each price was checked.

Connect

Server URL (read-only):

https://localaiprices.com/mcp

Claude Code:

claude mcp add --transport http localaiprices https://localaiprices.com/mcp

Apps that take a JSON config:

{
  "mcpServers": {
    "localaiprices": {
      "url": "https://localaiprices.com/mcp"
    }
  }
}

Can't use MCP? /llms.txt and /llms-full.txt give a language model the whole site as text.

Price alerts

Paste this into an assistant that can run on a schedule, fill in your models and budget, and it messages you when something you'd buy gets cheaper.

Check localaiprices.com for me each morning and tell me if a machine I'd buy got cheaper.

My models: <for example gpt-oss-120b>
My budget: <for example $4,000, or EUR 3,500>

Use the localaiprices MCP server (https://localaiprices.com/mcp). If you can't, use https://localaiprices.com/api/v1/changes?since=<time> and https://localaiprices.com/api/v1/models.json.

1. find_for_model for each of my models, within my budget. Those are my machines.
2. list_changes since your last run (first time: the last 7 days). Save next_since for tomorrow.
3. For each price drop, restock or new machine among my machines, write one line: name, old price to new price, store, tok/s on my model, date checked, localaiprices link.
4. Nothing relevant? Say "Nothing new today." and stop.

Don't buy or reserve anything. Don't guess prices or speeds; use only what the data says.

Plain text: /agent-prompt.txt. Not using an agent? Subscribe to the RSS feed or read Price changes.

Plain JSON

GET /api/v1/machines.json: Every machine: prices per store, check times, history summary, speeds and value.
GET /api/v1/skus.json: Every machine in one flat list, sorted by tok/s per $1,000.
GET /api/v1/changes.json: All price changes, newest first. Rules for what counts are in meta.rules.
GET /api/v1/changes?since=2026-10-20T00:00:00Z: Only changes after since (ISO date or time). Also takes type and limit (1 to 500).
GET /changes.xml: RSS: the 50 newest changes.
GET /api/v1/models.json: Models, memory needed, and which machines fit them with measured tok/s.
GET /api/v1/history/<id>.json: Every price observation for one machine.
GET /api/v1/prices/latest.json: The latest observation per store.
GET /api/v1/benchmarks.json: Benchmark sources (llm-speed part CC BY 4.0).
GET /api/v1/cloud.json: Cloud GPU hourly prices.
GET /llms.txt: Site overview written for language models.
GET /llms-full.txt: Overview plus the full ranking, one file.
GET /agent-prompt.txt: The price alert prompt as plain text.

Using it

Use it for anything, commercial too. Credit localaiprices.com with a link, show the check date next to prices, and keep the source link next to any speed. Checking once a day is enough, since that's how often prices update.