Adaption
Import, adapt and expand training datasets, then launch AutoScientist fine-tuning runs from an AI agent.
Query what Gen Z and Gen Alpha respondents say about brands in Claude.
Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.
Paste it into any MCP client. Setup docs
The Cafeteria MCP server lets brand and research teams ask Claude what young consumers think, answered from Cafeteria's own Gen Z and Gen Alpha survey responses. Its five tools rank the brands people mention, compare named brands by age, gender and state, and run read-only SQL. The only write sends Cafeteria a custom-research request, and only when you ask.
Verified connector
Listed by Anthropic as a partner connector in its Connectors Directory.
Connection checked by Agentman on .
Anthropic states this reflects the level of review a connector received, not a security audit.
Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.
query_cafeteria_tables returns at most 1,000 rows and is read-only.A table is Cafeteria's name for one research session: a focused set of questions on a topic, sent to participants through the Cafeteria app. Cafeteria Tables are always on and shared across all clients. Custom Tables are built for one brand, and Cafeteria says their answers are never shared outside that organization. The connector queries only the tables your organization is licensed for.
They are Cafeteria app participants aged 14 to 26, per Cafeteria's client FAQ. Everyone is invited or accepted from a waitlist, then quality-scored, and Cafeteria removes low-effort respondents. It says it benchmarks the base against U.S. census data by state and ZIP code. Respondents answer by voice note or text, so the connector returns their own words.
Cafeteria says broad tables exceed 385 responses, the common benchmark for 95% confidence and a 5% margin of error. For narrowly targeted custom studies it works with about 100 responses, which it calls decision-ready. A filtered query in Claude can cut a table well below either number, so check the unique-respondent count each tool returns before quoting a result.
https://mcp.cafeteria.page/mcp (2026-11-18): anonymous initialize returned HTTP 401 with WWW-Authenticate: Bearer and a resource_metadata URL. No tool was called. · retrieved 2026-11-18https://mcp.cafeteria.page/.well-known/oauth-protected-resource/mcp (2026-11-18). Source for the authorization server. · retrieved 2026-11-18docs/marketing/data/anthropic-directory-2026-11-17.json. Source for the tier, endpoint, five tool names and their descriptions. · retrieved 2026-11-17Connect Cafeteria once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.