Brian — Memory for Claude
Claude knows you exist.
Brian knows your work.
Claude's built-in memory holds your name and a few surface facts. Brian gives Claude a structured, searchable knowledge layer — your active projects, decisions, documents, and preferences — so every conversation picks up where the last one left off.
The problem
Claude's memory is a sticky note. Brian is a knowledge system.
Claude knows your name. Not your work.
Its memory holds surface facts — not your active projects, open decisions, or what you built last month.
You have no idea what Claude actually remembers.
There's no way to see it, edit it, or tell it what matters most. It's a black box you don't control.
Claude's memory has no concept of projects, documents, or decisions.
Everything lives in a flat, unsorted blob. No structure, no priority, no organisation.
How it works
Three things happen when Brian is connected
Step 01
Brian remembers
When you tell Claude something important, Brian stores it — tagged, scored, organised.
Step 02
Brian retrieves
Before every response, Brian finds relevant memories and injects them into context.
Step 03
Brian persists
New device. New conversation. Weeks later. Claude still knows.
The quality argument
Why Brian makes Claude smarter, not just faster
Most people assume memory tools are a convenience. Brian is also a quality improvement.
When you re-explain context manually, it bloats the conversation. Claude's attention gets spread across repeated background, pasted setup text, and accumulated noise — before it even reaches your actual question.
Research on large language models shows they perform worst on information buried in the middle of long contexts. Manual context management doesn't just waste your time. It degrades the output.
Brian works differently. It retrieves only what's relevant — typed, scored, and structured — and injects it cleanly before Claude responds. The context window stays lean. Claude's attention stays focused. The answers get better.
Less noise. More signal. Every time.
Selective retrieval
Only relevant memories enter the window. Not everything you've ever told Claude.
Typed and scored
Brian knows the difference between a fact, a preference, and an instruction. Better structure, better reasoning.
Clean context
No re-explanation noise. Claude works from a curated brief, not a bloated thread.
Context intelligence
Claude Projects floods the window.
Brian prunes the graph.
Claude's native Projects crams every file into the context window until it's full. Brian traverses the memory graph and sends only what's relevant to your specific request.
Claude Projects
Flat file dump
Upload 10 files. Claude tries to read all of them. The context window fills with noise until it can't hold any more.
~94k tokens used
10/10 files injected
low signal ratio
Brian MCP
Graph traversal
Brian reads your query, traverses the memory graph, and injects only the nodes relevant to your specific request.
~5k tokens used
5/47 nodes selected
high signal ratio
Memory architecture
Every memory has a lineage.
Brian doesn't store isolated facts. Each memory is tethered to the conversation, document, and decisions that created it. Hover a node to explore its connections.
Capabilities
Everything Claude needs to know you
Memory spaces
Separate contexts for every project. Switch between work, personal, and client projects without cross-contamination.
Google Drive sync
Index documents from Drive directly into memory. Claude can reference your specs, reports, and files.
Smart retrieval
Relevance scoring, salience weighting, and semantic search surface the right memories for every conversation.
Pin what matters
Pin critical memories so they're always in context. Your most important instructions and facts, every time.
Works everywhere
Claude web, Claude desktop, Claude mobile. One memory, every device.
Full control
Store, edit, delete, organise. Your memory, your rules. Nothing is hidden.
Use cases
Built for people who use Claude every day
Track a project
Set up a space for your project. Brian tracks every decision, milestone, and open question across sessions.
Never re-explain
Your role, your stack, your preferences — stored once, available forever.
Reference your documents
Connect Google Drive. Brian indexes your specs, reports, and files for Claude to reference anytime.
Session continuity
Pick up exactly where you left off. Brian replays the context so Claude doesn't miss a beat.
Multi-project context
Working on three things? Brian keeps them separate. Switch spaces, switch context.
Extract insights
Brian pulls key decisions and facts from your conversations automatically.
Give Claude a brain worth trusting.
Install Brian MCP and stop managing context manually.
Get started free →