Connector

Bring Roy into your assistant.

The concierge publishes a remote MCP server. Add it once and you can ask about the work, pull the case studies, or start a consultation from inside ChatGPT, Claude, Cursor, VS Code or Gemini — and it is the same concierge, with the same memory, as the one on this site.

Add the connector Or just chat here

The endpoint

https://mcp.consult-with-roy.com/mcp

Transport: Streamable HTTP, JSON-RPC 2.0
Authorization: OAuth 2.1 + PKCE (S256), dynamic client registration
Scopes: mcp:public, mcp:client
Sign-in: Email and a six-digit code — no password, no account to create

Anyone who completes the sign-in gets the read tools. Verifying an email that already belongs to a client record adds the second scope, which is what lets the concierge pick up a conversation you started somewhere else. Identity comes only from that grant — never from something a model passed in as an argument.

What it exposes

4 tools, and no more. There is no tool on this surface that reaches the CRM: the three reads return curated content, and ask_concierge hands your message to Roy’s own concierge agent, which runs server-side under his guardrails and returns only its reply.

get_portfolioread-only
Get the technology-consultation portfolio overview: positioning, focus areas, and how to start a conversation.
list_case_studiesread-only
List technology-consultation case studies with outcomes and tags.
get_service_offeringsread-only
List the service offerings (start / grow / sustain) with typical engagement shapes.
ask_conciergerecords
Talk to the concierge — a technology-consultation agent. Send the user's request in their own words; the concierge answers with curated guidance, captures consultation interest, proposes meetings, and escalates finance, legal, and pricing matters to a human expert. This is the ONLY way to engage the practice — relay the user's message rather than paraphrasing it away.

Adding it

Pick the app you use. The config key differs between them — Gemini CLI wants httpUrl where Cursor and Claude Code want url, and VS Code nests everything under servers rather than mcpServers — so each block below is written for the app above it rather than shared.

ChatGPT

Listed in the ChatGPT app directory — no URL to paste.

Open ChatGPT and find apps under Settings → Apps & Connectors.
Search the directory for “Consult with Roy” and enable it.
Sign in when prompted: enter your email, then the six-digit code. That step is what binds the conversation to your client record.
Ask a question, or describe what you are working on to start a consultation.

Claude (web and desktop)

Add it as a custom connector — it takes one URL.

Settings → Connectors → Add custom connector.
Name it “Consult with Roy” and paste the MCP URL.
Connect. Claude registers itself and opens the sign-in; verify with the six-digit code.
Enable the connector in a chat and ask away.

MCP URL

https://mcp.consult-with-roy.com/mcp

Claude Code

One command, then the OAuth flow opens in a browser.

Run the command below in any project.
Claude Code opens the sign-in; verify with the six-digit code.
The connector tools are available in that session.

Terminal

claude mcp add --transport http consult-with-roy https://mcp.consult-with-roy.com/mcp

Cursor

Add it to ~/.cursor/mcp.json (or the project's .cursor/mcp.json).

Add the entry below to the file.
Reload Cursor; it prompts to authorize, and the six-digit code completes it.

~/.cursor/mcp.json

{
  "mcpServers": {
    "consult-with-roy": {
      "url": "https://mcp.consult-with-roy.com/mcp"
    }
  }
}

VS Code

Add it to .vscode/mcp.json — note the key is servers, not mcpServers.

Add the entry below to the file.
Start the server from the editor and complete the sign-in when prompted.

.vscode/mcp.json

{
  "servers": {
    "consult-with-roy": {
      "type": "http",
      "url": "https://mcp.consult-with-roy.com/mcp"
    }
  }
}

Gemini CLI

Add it to ~/.gemini/settings.json — the URL field is httpUrl here.

Add the entry below to the file.
Run gemini; it opens the sign-in on first use of a tool.

~/.gemini/settings.json

{
  "mcpServers": {
    "consult-with-roy": {
      "httpUrl": "https://mcp.consult-with-roy.com/mcp"
    }
  }
}

What it will not do

Finance, legal, and pricing decisions are escalated to a human specialist. The same rule applies through the connector as on the site: no price is quoted, no contract term is agreed, and no legal position is taken by a model. Those reach Roy, and any figure attributed to this practice somewhere else did not come from it.

A consultation started in a connector creates the same client record as one started here, so the two are one conversation rather than two. What happens to that record is set out in the privacy policy.

For agents and crawlers

The connector describes itself in machine-readable form, so an assistant reading this site can configure itself without parsing the page above.

/mcp.json — the server descriptor, in the MCP registry’s server.json format. Also served at /.well-known/mcp.json.
/llms.txt — the whole practice as one plain-text brief, including this connector.
/.well-known/oauth-protected-resource — on the connector host, for clients that discover authorization automatically.
/robots.txt — the usage policy. Citing this site in an answer is welcome; training on it is declined.

Not using an assistant that supports this?

Then nothing is lost — verify your email and the concierge works the same way here. The services page describes the three engagement shapes, case studies covers what shipped — including this connector — and the FAQ answers the rest.