Ashley Payne, fractional CTO

Ask a UK fractional CTO's practice what it does, what it costs and whether it suits your project. Read-only, no sign-in.

View on ashleypayne.co.uk

Hosted MCP Server

npx add-mcp 'https://www.ashleypayne.co.uk/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

Ask an assistant about me.

If you use Claude, or Codex in the ChatGPT desktop app, you can connect my business to it and ask about my case studies, prices, writing and what I’m wrong for, in your own words. I never see the conversation, nothing your assistant sends my server is stored, and the privacy page says exactly what reaches me.

Connecting it

Steps for Claude, and for Codex in the ChatGPT desktop app.

Claude

  1. Open settings and go to Connectors.
  2. Choose Add custom connector.
  3. Name it Ashley Payne and paste the address above.
  4. It’ll say it can’t check the server, or can’t work out how it signs in. There’s nothing to sign in to, so choose Continue anyway.
  5. Turn it on in a chat from the tools or + menu.

Works on Free, Pro, Max, Team and Enterprise, no developer settings needed (Free is limited to one custom connector).

Codex in ChatGPT desktop

  1. Open the desktop app and select Codex in the ChatGPT dropdown.
  2. Open Settings, then MCP servers and Add server. In versions with a Plugins menu, choose Add, then Add MCP server.
  3. Name it ashley_payne, choose Streamable HTTP and paste the address above as the URL.
  4. Leave the bearer token and header fields empty.
  5. Save, then select Restart.
  6. Start a new Codex chat and choose Local.
  7. Type /mcp to check ashley_payne is connected, then ask your question.

This connects Codex running locally in the ChatGPT desktop app. It doesn’t make the server available in ordinary ChatGPT chats or in the browser. The Codex CLI and IDE extension can use the same connection on this computer.

Ask it the questions you wouldn’t ask me.

I built it to tell you when I’m the wrong choice, and if your wording is unclear it won’t guess. It asks your assistant three plain questions, and the rule that turns the answers into a yes or a no is mine and published. Your assistant passes the answers on and doesn’t decide what they mean.

  • Is Ashley the right fit for what I’m building?
  • What would working with him cost?
  • Has he built anything like my problem before?
  • How does he decide whether AI is the right answer?
  • What would he tell me not to do?

If it says I’m not a fit, take that seriously, but it only knows what your assistant passed on, and I can’t see whether that came from you or something it inferred. If the answer surprises you, tell me what it said and I’ll give you mine.

I build these for a living.

When staff use an AI assistant with no access to company context, it invents one. I build MCP servers over real systems to fix that, and this site runs on the same thing I’d build for you.

The coding education case study shows the same pattern over live booking and pipeline data.