ChainReaction for AI assistants
Ask Claude, ChatGPT or any assistant that speaks MCP what's good at any chain, and get our numbers back — the CR Score, the rating count and a link — instead of a guess.
MCP server
https://chainreactionhq.com/api/mcp
Streamable HTTP · read-only · no API key · free
Add it
- Claude — Settings → Connectors → Add custom connector, paste the URL above. No sign-in step: it connects straight away.
- Claude Code —
claude mcp add --transport http chainreaction https://chainreactionhq.com/api/mcp - ChatGPT, Cursor and other MCP clients — add a remote MCP server and paste the URL. Anything that supports Streamable HTTP without authentication works.
Then ask
- “What's the best thing to order at Culver's?”
- “Whopper or Big Mac — which do people actually rate higher?”
- “Which chain has the best fries?”
- “What's the worst-rated item at Taco Bell?”
- “Which fast food chain has the best menu overall?”
Every answer carries the rating count and a link to the item, chain or report page, so you can check it. The assistant is told to say when an item has too few ratings to call.
The tools
search_items— find items by name, optionally within one chainget_item_score— one item's CR Score, tier, rating count, star average and rank in its chaincompare_items— two items head to head, from any chains — scores, winner, margin, comparison pagechain_rankings— every item at one chain ranked, plus the chain's standing against other chainsbest_in_category— the top items in a category — fries, nuggets, chicken sandwiches, desserts — sitewide or at one chaincensus_summary— the 2026 Fast Food Census in numbers: chains ranked, best and worst items, CSV link, how to cite
What the numbers are
The CR Score is the share of an item's ratings at 3.5 stars or higher, smoothed toward the sitewide average and shown 0–100. Fire is 75 and up, Solid 50–74, Skip under 50. An item needs 5 ratings to be scored and 25 to be ranked against items from other chains; a chain is ranked against other chains once 80% of its catalogued menu has a score. Ratings come from paid survey panels of U.S. consumers, members and anonymous visitors. How the score works →
The server answers from a snapshot of the catalog refreshed hourly, so it is never more than an hour behind the site. It is the same data as the pages — nothing is available here that isn't public there.
Terms
Free to use, including commercially, with attribution and a link to ChainReaction (CC BY 4.0 — the same licence as the 2026 Fast Food Census). Scores change as ratings arrive; quote the rating count with the score and the date you pulled it. There is no rate limit today; if you are building something that will call it heavily, say hello first.
Other ways in
- The whole dataset as a CSV: /reports/fast-food-census-2026/data.csv
- A plain-text guide to the site for language models: /llms.txt
- schema.org structured data on every item, chain, ranking and comparison page.
Contact
hello@chainreactionhq.com — for a chain you want scored, a tool you wish existed, or anything the server gets wrong.
