Claude is genuinely useful for business work — drafting, analysis, research, writing. But there's a problem every heavy user hits: Claude starts from scratch every conversation. No memory of your clients, your ongoing projects, or the decisions you made last month.
That friction compounds fast. You spend the first few minutes of every session re-briefing. Or you don't, and Claude gives you generic answers because it doesn't know your situation.
This is solvable. Here's how business users manage it.
Who the client is, their industry, their constraints, past decisions, communication preferences. Every time you start a new client session, you rebuild this from scratch — or you get generic output.
What's been decided, what's in flight, what the constraints are. Projects live across weeks or months. Claude's session-based context can't hold that.
Why you made the calls you made. This matters for consistency and for onboarding new people. Decisions evaporate unless you record them somewhere Claude can reach.
Seniority, domain, preferred output format, tone. Claude defaults to generic unless you tell it your context.
Works, but it's friction. You need to maintain a doc of your context and remember to paste it. For busy periods with multiple clients, this falls apart quickly.
Claude's Projects feature lets you attach up to 5 documents that Claude has access to throughout the project. Better than nothing — but one project per context, 5-file limit, no search, no structured data.
Connect a structured store via MCP. Claude loads what it needs — your role context, client records, current projects — with a single call. The store is queryable, not a flat document. It scales across dozens of clients and projects without hitting file limits.
Here's a typical structure for a consultant or freelancer:
Collections:
context — your role, expertise, working style (loaded every session)clients — one record per client: industry, constraints, preferences, statusprojects — active engagements: what's in scope, key decisions, next actionsdecisions — a running log of significant calls you've madecontacts — key people with notes on relationship + last interactionDaily use:
User: context()
Claude: [loads your standing context — role, current projects, active clients]
Ready. What are we working on?
User: I'm preparing for the Acme quarterly review. What do we have on them?
Claude: [searches clients + projects for "Acme"]
Here's what I have on Acme Corp: [summary of client record]
Current project: Q3 strategy review. Last decision logged: chose Option B on pricing.
User: Good. Draft the exec summary.
[Claude writes knowing the client, the project, and past decisions]
Compare that to a blank-context session where you spend 5 minutes briefing Claude before you can get to actual work.
One of the most useful patterns for heavy Claude users: start each morning with a context load that primes Claude for the day's work.
context(expand=["clients","projects","decisions"])
This loads your standing context plus your active client records, live projects, and recent decisions into a single call. Claude arrives fully briefed. You go straight to work.
The free tier (2,500 records, 50 searches/month) covers most individual business users. Pro tier (100,000 records, 1,000 searches/month) at £8/month is for heavier use or larger record sets.
If you spend 5 minutes re-briefing Claude per session and use Claude 3 sessions a day, that's 15 minutes a day, ~5 hours a month. Stash pays back in a week.
The setup takes 10 minutes. The payback is every session after that.
Add Stash to Your Business Claude Workflow
2,500 records free. No more re-briefing. Connect in 30 seconds.
Get Started →