Every few months, an AI update lands that actually changes how you work — not just a shinier interface or a marginal speed boost, but something that removes a real daily frustration. Anthropic just shipped two of those updates at once, and most people have no idea how to use either of them.
Here's what changed: Claude now remembers your preferences, context, and working style between sessions. And separately, every Claude user — including free-tier accounts — now has access to a one-million-token context window. That's roughly 750,000 words in a single conversation. If you've been re-explaining your business to Claude every time you open a new chat, or hitting walls when you try to paste in a long document, both of those problems are now solved.
What "persistent memory" actually means
Until recently, every Claude conversation started from zero. You'd open a new chat, explain that you run a 12-person marketing agency, that your clients are in the finance sector, that you prefer bullet-point summaries over long prose — and then you'd do it all again tomorrow. That friction wasn't huge, but it added up.
Persistent memory changes this. Claude can now store facts about you and your preferences across sessions. Tell it once that you hate jargon, that your business uses a specific naming convention, or that you always want responses formatted for a non-technical audience — and it remembers. Anthropic also added a memory import tool, so if you've been building up context in ChatGPT or Gemini, you can migrate that stored knowledge directly into Claude without starting from scratch.
This isn't just a convenience feature. For anyone using Claude as a regular work tool, it shifts the relationship from "disposable assistant you brief every time" to "context-aware collaborator that already knows your situation."
What a million-token context window unlocks
Context windows are the amount of information Claude can hold in mind during a single conversation. A million tokens is enormous — enough to load an entire company handbook, a year's worth of email threads, or a lengthy legal contract (or several) in one session.
Previously, long-context capability was reserved for higher-tier plans or enterprise access. Now it's available across all Claude tiers, including Haiku and Sonnet. That means cost-sensitive workflows — the kind that SMB owners actually run — can now use long-context without paying enterprise pricing.
The practical shift here is significant. You're no longer constrained to working with excerpts or summaries of your documents. You can bring the whole thing in and ask real questions about it.
Four use cases worth trying this week
These aren't hypothetical. These are workflows that save real time, and both upgrades make them meaningfully better.
1. Onboarding documentation
Paste your entire onboarding pack — role descriptions, team structure, tool guides, culture notes — into a single conversation. Then ask Claude to identify gaps, inconsistencies, or anything a new hire would find confusing. With a million-token window, you're not cherry-picking which sections to include. The whole doc goes in. Memory means you can return to the same conversation thread the next day and Claude still knows you're building for a junior hire joining the operations team.
2. Contract and policy review
Upload a full vendor contract or policy document and ask Claude to flag non-standard clauses, summarise obligations, or compare it against a template you've pasted alongside it. Long documents that used to require chunking or a specialist tool now fit comfortably in one session. This isn't legal advice — but it's a solid first pass before you pay a lawyer to read the whole thing.
3. Ongoing project context
If you're managing a multi-month project — a website rebuild, a product launch, a system migration — persistent memory means Claude can hold the thread. Set it once: project goals, stakeholders, constraints, current status. Then use it as a thinking partner across weeks without repeating yourself. It's closer to how you'd work with a consultant who actually reads their notes before the next meeting.
4. SOP drafting from messy notes
Most standard operating procedures exist in someone's head, or scattered across Slack threads, handover emails, and half-finished documents. Dump everything into one conversation — all of it, however unstructured — and ask Claude to produce a clean, step-by-step SOP. The million-token window means you don't have to curate the input. Memory means you can iterate on the draft across sessions without re-uploading the source material each time.
How to get the most out of memory
Memory is only as useful as what you put into it. A few things worth doing immediately:
- Set your working style explicitly. Tell Claude how you prefer responses formatted, what level of detail you want, and any terminology that's specific to your business.
- Store your business context. Industry, team size, key products or services, current priorities. The more Claude knows upfront, the less prompting every individual task requires.
- Review what's stored periodically. Memory isn't magic — if your business context changes (new focus, new team structure), update what Claude knows. Stale memory is worse than no memory.
If you're already using Claude regularly and want a broader sense of how it fits into your tool stack, the post on choosing the right AI assistant for your business covers how Claude compares against the alternatives and what it's genuinely better at.
The bigger picture
The race for longer context and persistent memory has been running in the background of AI development for a while — it's what separates a useful tool from a genuinely capable collaborator. What's notable here isn't just that Claude got these features. It's that Anthropic made them free, across all tiers, right now.
That's a meaningful shift for SMB owners who've been running on free or entry-level plans and hitting walls. The workflows that used to require an enterprise plan or a custom integration are now available without a procurement process or a budget conversation.
The best AI tools don't feel like tools — they feel like working with someone who already knows your situation. Persistent memory and long context are how that starts to happen.
You don't need to overhaul anything. Pick one of the use cases above, try it this week with your actual documents, and see what the new context window changes about how you work. That's the fastest way to find out whether this upgrade is as useful as it looks — and in this case, it probably is.
Sources
This article is grounded in the following reporting and primary-source announcements.