connect · MCP server
My CV has an interface for your AI.
Add this site to Claude, or any client that speaks the Model Context Protocol, and ask your AI about my work. It answers from the same published record every page here is written from, as of 8 October 2026. Read-only, no sign-up, nothing stored about you beyond a daily count per tool.
The address
https://andrewdeighan.com/mcp
Streamable HTTP, no authentication. It accepts POST only and never opens a stream back to you.
Add it to Claude
- In Claude (web or desktop), open Settings, then Connectors.
- Choose Add custom connector.
- Name it Andrew Deighan and paste the address above as the remote MCP server URL. Leave the OAuth fields empty.
- Save. In a new chat, open the tools menu, make sure the connector is switched on, and ask a question.
Custom connectors are available on Claude's paid plans. An organisation admin may need to allow them first.
Other MCP clients
Any client that supports remote servers over Streamable HTTP can use the same address with no credentials: for example Cursor, Claude Code (claude mcp add --transport http andrew https://andrewdeighan.com/mcp), or the MCP Inspector.
Three questions to try
- What has Andrew built with AI agents, and what were the results?
- How does Andrew keep AI systems under control, and what happens when a model fails?
- What evidence is there for the 1% to 25% reply-rate figure?
What your AI can call
| Tool | Returns |
|---|---|
| get_profile | Returns the published profile: name, title, availability, location, right to work, contact email, links, roles, background and known gaps. |
| list_projects | Returns the published list of projects, one line each, with ids for get_project. |
| get_project | Returns one project in full: problem, what was built, how the AI is controlled, measured results, and the page URL. Each line cites fact ids. |
| get_evidence | Returns one fact from the fact file by id: its text, the figures in it, the project it belongs to, the page it is shown on, and the record date. |
| search_record | Returns the facts, project lines, profile entries and evaluation summary whose text contains the query words, most matches first. |
Everything returned is data from the fact file and the project pages, generated at build time from the same source as this site. The server calls no model, and a build-time check fails if any output could read as an instruction to the AI calling it. 48 facts, 4 projects, as of 8 October 2026.