AI & Agent Access (MCP)

AI agents and developers can query my approved public professional profile, projects, publications and research evidence through a read-only Model Context Protocol server. Results are structured, source-linked and carry explicit evidence classifications and limitations.

The service performs deterministic retrieval only. It does not generate answers, fetch arbitrary URLs or provide access to Calixa's model-backed services. Private, unlisted, password-protected, NDA-protected and withdrawn material remains unavailable.

Endpoint: https://roblemumin.com/mcp

Streamable HTTP · POSTMCP 2026-07-28Public · read-onlyNo inference

An MCP-compatible client is required. Opening /mcp in a browser does not start a conversation and may correctly return HTTP 405.

Supported tools

  1. get_public_profile
  2. list_capabilities
  3. list_projects
  4. get_project
  5. get_public_career
  6. list_public_evidence
  7. search_public_evidence
  8. get_public_evidence
  9. list_publications
  10. get_publication

What agents can retrieve

  • Public profile, career, skills and credential summaries.
  • Project descriptions with maturity boundaries.
  • Publications, DOI references and review status.
  • Ranked public evidence with matched source passages.
  • Bounded evidence chunks with reconstruction hashes.

Evidence boundaries

  • Published by the author is not independent verification.
  • An implementation is not evidence of production adoption.
  • Internal tests are not independent certification.
  • A research proposal is not an experimentally validated theory.

List the tools

This request was tested against the deployed protocol shape. MCP mirrors the method and target into HTTP headers.

curl -sS https://roblemumin.com/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json' \
  -H 'MCP-Protocol-Version: 2026-07-28' \
  -H 'Mcp-Method: tools/list' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{"_meta":{"io.modelcontextprotocol/protocolVersion":"2026-07-28","io.modelcontextprotocol/clientCapabilities":{},"io.modelcontextprotocol/clientInfo":{"name":"public-example","version":"1.0.0"}}}}'

Retrieve the public profile

curl -sS https://roblemumin.com/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json' \
  -H 'MCP-Protocol-Version: 2026-07-28' \
  -H 'Mcp-Method: tools/call' \
  -H 'Mcp-Name: get_public_profile' \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"get_public_profile","arguments":{"language":"en"},"_meta":{"io.modelcontextprotocol/protocolVersion":"2026-07-28","io.modelcontextprotocol/clientCapabilities":{},"io.modelcontextprotocol/clientInfo":{"name":"public-example","version":"1.0.0"}}}}'

The structured result identifies the returned language, canonical profile, evidence status, limitations and projection hash. Publication and evidence tools follow the same source-bound pattern.

Discovery

Domain-level MCP discovery remains experimental and is not supported automatically by every client.

Security and limits

  • No login is required for eligible public records.
  • No write, upload, deployment or administrative operation exists.
  • No paid model, embedding or external-fetch operation exists.
  • Requests, pages and document chunks are size-bounded.
  • Unavailable or unknown records return explicit outcomes; no substitute is invented.