MCP-compatible access · self-serve or managed

Irish grocery intelligence your AI analyst can actually use.

BasketWatch Agent Access exposes the live grocery dataset through a managed private layer for analyst copilots, internal chat tools and automated reporting workflows. It is designed for commercial teams that want answers with prices, promotions, unit prices and evidence attached.

Self-serve or provisioned Usage limits by plan Evidence-backed answers Agents can buy their own access

Two ways in

Connect it yourself, or have it provisioned.

Most teams do not need anything bespoke. The MCP server runs locally against your own API key, so your assistant queries the live dataset directly and your credits meter exactly as they do on a direct HTTP request. Nothing proxies through us.

Connect it yourself

{
  "mcpServers": {
    "basketwatch": {
      "command": "python",
      "args": ["-m", "basketwatch_mcp"],
      "env": {
        "BASKETWATCH_API_BASE": "https://basketwatchireland.com",
        "BASKETWATCH_API_KEY": "YOUR_KEY",
        "BASKETWATCH_MCP_MAX_ROWS": "200"
      }
    }
  }
}

Seven tools: search, cross-retailer comparison, promotions, price changes, new lines, removed lines and catalogue listing. Each call returns up to 200 rows by default, and without a key you still get a live 50-row sample, so you can try it before buying. Full setup is in the API reference.

Your credit pack covers this. No separate MCP plan and no second balance: a pack bought on the pricing page is the same credits whether the records are pulled by your own code or by your assistant. One record returned, one credit.

Let the agent buy its own

An autonomous agent can purchase its own credits and start querying, with no browser and no card entry, by paying over MPP with a Stripe shared payment token.

  1. Reads the price at /api/credits/mpp
  2. Receives a signed 402 payment challenge
  3. Pays, and gets back a funded API key

Payment terms are HMAC-bound, so an edited challenge is refused, and the charge is idempotent, so a retry cannot bill the card twice. Documented in the API reference and described for machines at /llms.txt.

Managed access still exists for teams who want scoped workflows, dedicated capacity or an SLA. That is a conversation, not a signup.

Live agent session

Question in, evidence out , watch it work.

A real workflow: a question goes in, the agent calls structured BasketWatch tools over MCP, and the answer comes back with retailers, prices and unit prices attached, never a guess.

basketwatch-agent · MCP session live
› ↵ ask

What it can do

A grocery analyst layer on top of the BasketWatch feed.

The agent layer is not a general chatbot. It is a focused interface over structured Irish grocery data across Aldi, Tesco, SuperValu and Dunnes Stores.

Compare products across retailers

Ask for a product, pack size or category and return matched product groups with shelf prices, member prices and unit prices across the four major retailers.

Find current promotions

Surface current Clubcard, Real Rewards, multibuy, was-price and member-only offers with the underlying offer text preserved.

Explain price movement

Turn daily changes into readable summaries: what rose, what fell, by how much, and where the evidence came from.

Watch range changes

Identify newly listed products, removed products and range churn by retailer, category or search term.

Generate analyst briefs

Build daily category notes, promotion roundups, competitor snapshots and customer-ready summaries from the latest snapshot.

Keep answers auditable

Responses can include scrape dates, retailer names, product URLs, IDs and row-level context so an analyst can verify the conclusion.

Private tool layer

Structured tools, not scraped answers.

The agent uses controlled BasketWatch data tools, then writes the answer from returned rows. That keeps the LLM away from guessing prices or inventing promotional context.

search_products compare_price_across_stores lookup_barcode get_promotions recent_price_changes newly_added_products removed_products status

What comes back with the answer

Product identity

Retailer, product name, product ID and product URL where available.

Pricing

Shelf price, member/effective price and unit price basis.

Promotions

Promotion label, was-price, loyalty pricing and validity text where captured.

Freshness

Scrape date and freshness context for the data used.

No guessing

Clear “not enough data” responses when a match is weak or unavailable.

How access works

Provisioned privately for commercial use.

The public site shows what the layer can do; the implementation stays private so access is controlled, metered and supported. Four steps from scope to live.

Step 1

Scope the workflow

Agree the use case: analyst chat, daily brief, watchlist, comparison or internal copilot.

Step 2

Set commercial limits

Usage caps, data access, cadence and support level matched to the team and subscription.

Step 3

Provision access

Approved customers get private setup details and keys. Connector internals stay unpublished.

Step 4

Review outputs

Answers check against row-level evidence so teams trust the data behind the summary.

Pricing model

Sold as a direct commercial add-on.

Separate from the public API marketplace. Best for teams that already know they want BasketWatch data inside an AI workflow and need controlled access, support and usage boundaries.