Whether you are breaking into defence, delivering a contract, or working to stay competitive, Alexandria answers your delivery and product questions from MilUX’s own record of doing the work: what actually goes wrong, the tells to look for, and what to do first. When it does not know, it says so.
Alexandria is an MCP server. Connect it once and MilUX judgement is available inside the assistant you already work in, so you can check a claim while you are writing the bid paragraph rather than after you have sent it. No new tab, no separate login.
It connects to anything that speaks MCP: ChatGPT on any paid plan, through Developer Mode in settings. Claude on every plan, including free. It also works in Claude Code, Cursor, VS Code with Copilot, Perplexity and Mistral Le Chat. Free ChatGPT accounts cannot add custom connectors, and on a Business or Enterprise workspace your administrator may have turned Developer Mode off.
https://alexandria.milux.co.uk/api/mcp“What decides whether an MOD contract needs Cyber Essentials or Cyber Essentials Plus, and where is that stated?”
The determinant is the cyber risk level assigned to the contract, not the size of the supplier or the value of the work. Contracts carrying DEFCON 658 require compliance with Def Stan 05-138; Cyber Essentials is the floor for the lower tiers, and Cyber Essentials Plus becomes mandatory once the contract sits at moderate risk or above. The tender documents should name the risk level. If they do not, ask the buyer rather than guessing. This is one of the checks worth running at the pre-bid gate, before any bid effort goes in. bid-0010
A real answer, abridged. Every claim cites the note it came from.
In plain English, the way you would ask a colleague. In the web chat, or from your own AI tools over MCP.
Answers come from a curated corpus of real delivery work in defence, and cite the exact notes they draw on.
Where the corpus runs out, Alexandria says so and offers a fifteen-minute call with MilUX. It never improvises.
You never name a tool. Describe the situation and Alexandria runs the right method, or answers as the right specialist, from MilUX’s own practice. A single call is a read; the full end-to-end workflow behind a method is the engagement.
Most assistants answer everything, which is exactly why you cannot trust them on anything. Alexandria answers from one practitioner’s curated corpus, cites the note every claim came from, and when your question falls outside it, tells you plainly and offers a call instead. An honest “we do not cover that” is worth more than a confident guess, and it is the one thing a general model will never give you.
Prices exclude VAT. Each plan includes a monthly usage budget. You are billed by what each answer actually uses, and you can check how much of your budget is left at any time. Unused budget does not roll over. Registration takes a minute, no card. VAT-registered businesses enter their VAT number at checkout and are charged according to their status.
£0
£2 of usage a month to try
£5/month + VAT
An eighth of Standard's budget, at the same rate per question
Most popular
£40/month + VAT
Quick questions cost pennies; agent and instrument runs use more
£80/month + VAT
Double the budget, for a small team
£1,200/month + VAT
Generous usage, plus two consulting days
Invoiced. Limited to ten organisations
You are billed by the tokens each answer uses, so a quick question costs a few pence and a full instrument or agent run costs more. There is no separate credit currency to decode; you can check how much of your budget is left at any time, on your account page or with the usage tool. On paid plans that budget is shared across your team: you add colleagues by email, so no one hands round a key or a link.
Alexandria runs on Anthropic’s commercial API. Nothing you ask it, and nothing you upload to it, is used to train a model. A personal ChatGPT or Claude account can work differently: consumer terms may allow your conversations to be used to improve the provider’s models unless you have opted out. If you paste bid text or pricing into an assistant, that is worth knowing.
We do not store the text of your questions in any database or log. We keep a count for billing and an anonymous record of which topics come up, unlinked to you, so we know what to write next.
If you connect Alexandria to your own assistant, your question passes through that provider as well as through us, under whatever terms you hold with them. We can only speak for our side of the connection. If the work is commercially sensitive, connect from a business account rather than a personal one.
Alexandria is not accredited for government-classified material. Do not upload anything carrying a government classification.