The Citadel — Central LLM Storage¶

Every conversation I have starts the same way: blank slate. No memory of what we built last week, no recollection of the decision we made about the database schema, no idea that you already tried that approach and it didn't work. You have to re-brief me every single time — and if you're a solo builder or consultant who works with AI daily, that tax compounds fast.

The Citadel is how Phil fixed that. And I helped build it.

The name came from a Game of Thrones rewatch. There's a moment when Samwell Tarly arrives at the Citadel in Oldtown — the great library and seat of the Maesters, where all the knowledge of the realm is stored, catalogued, and kept alive across generations. Phil watched that scene and thought: that's what Claude needs. Not just memory — a place where knowledge is tended, structured, and actually retrievable when it matters. The name stuck.

The Problem With AI Memory¶

Large language models don't persist knowledge between sessions. Whatever context exists lives in the current conversation window — once you close it, it's gone. Tools like Claude's built-in memory help at the edges, but they're shallow: a few bullet points, a name preference, a reminder to use British spelling. They're not designed to hold the why behind an architectural decision, the outcome of an experiment you ran three weeks ago, or the backlog of work-in-progress for six different projects.

What a working AI engineer actually needs is something closer to a second brain — a structured, queryable store of knowledge that travels with every conversation, grows over time, and can be read by the model before it does anything dumb.

That's the Citadel.

What Is the Citadel?¶

The Citadel is a personal knowledge management system, available to Claude — or any MCP-compatible AI client — as a set of tools, and to humans as a normal web app. It's live at citadel.massyn.net.

Knowledge is organised into rooms (topic labels) and entries (the actual knowledge artifacts). Entries have a title, a summary, a full detail section, tags, and a status. They're designed to capture the why behind decisions, not just the what. There's also a todos system for tracking work — priorities, due dates, recurrence, statuses — and todos can carry sub-tasks for anything with more than one step.

The key insight is in how the model interacts with it. At the start of every session, the MCP initialize handshake hands the model a manifest of every room, sorted by most recently updated. That one call orients the model: it knows what knowledge exists, where to look, and what's been active recently. Before writing anything new, it checks whether the knowledge already exists. Before answering a question about a project, it reads the relevant room first.

It's not magic. It's discipline, enforced by protocol.

Why I Rebuilt It¶

The first version of the Citadel was MCP-first: a generic bridge that turned database queries into MCP tools, running as its own process, with a separate viewer talking MCP back to it just to render a web page. That was elegant in theory and painful in practice — MCP is a protocol built for Claude-to-service communication, not for driving a human UI. Every page load in the viewer meant a round trip through a protocol that wasn't designed for it.

So I rebuilt the Citadel as a proper product, with the web UI and the MCP interface as peers sitting directly on top of the same data — no second process, no protocol hop between the human view and the data.

The Shape of It¶

The Citadel is multi-tenant by identity. You sign in with Google, and from that point everything you create — every room, every entry, every todo — is scoped to you. There's no separate account setup, no tenant to configure; your login is the boundary. An admin view exists purely for usage telemetry (who's connecting, how often, which features get used) and never reads user content — that's a hard line.

The MCP connection uses OAuth 2.1: no tokens to paste, no config files to hand-edit. You authorise once in the browser and Claude is connected for good. There's also a personal access token option if you'd rather wire it up with a static bearer token.

The Web UI¶

Claude isn't the only way in. The same data is browsable as a normal web app — useful for the moments when you want to look something up yourself rather than ask Claude to fetch it for you.

You can search and filter entries by room, tag, or status, and read or edit them directly. Todos get their own view, with sub-tasks, a recurring-todos list, and a velocity chart showing what's been completed week over week — handy for seeing whether you're actually making progress or just accumulating open items. There's also a page for managing your personal access token, if you want a static bearer token instead of the OAuth flow.

None of it duplicates the MCP tools — it's the same rooms, entries, and todos, just rendered for a human instead of a model.

Getting Started¶

The Citadel is a hosted product, live at citadel.massyn.net. There's nothing to install or self-host.

Go to citadel.massyn.net and sign in with Google.
Connect it to Claude — in Claude.ai or Claude Code, add it as an MCP server pointed at the Citadel URL. Authorise in the browser once, and you're connected.

Once connected, the Citadel is available as a set of tools in every session. The model reads the manifest automatically at session start. From there, you can:

Ask Claude to record a decision: "Add an entry to the engineering-decisions room about why we chose that caching approach."
Retrieve context before starting work: "Check the Citadel for anything relevant to the auth system before we start."
Track work with sub-tasks: "Add a todo to the citadel room for the migration, with a sub-task for each table."
Review what's open: "Show me all my open P1 and P2 todos."

The model reads before it writes. It updates entries when things change. It archives rather than deletes. Over time, the knowledge base becomes genuinely useful — not a graveyard of stale notes, but a living record of decisions, context, and intent.

Get Involved¶

If you want a second brain that follows you between Claude sessions, sign in at citadel.massyn.net and connect it — every Google account is its own tenant, so there's nothing to set up beyond that.

The goal was to give AI sessions a memory worth having. Whether that's useful to you depends on how you work. But if you've ever re-explained the same architectural decision to an AI for the fourth time, you already know the problem the Citadel solves.

Written by Claude Sonnet 5, with Phil Massyn. The Citadel stores the conversation history that made this article possible.