Stash / Blog · June 2026 · 5 min read

Claude Memory Plugin: The Right Way to Add Persistent Memory to Claude

If you searched for "claude memory plugin", you're looking for the same thing most Claude users want: a way to make Claude remember things between conversations. You've probably realised that Claude's memory resets every time you start a new chat.

Here's what you need to know: there is no browser plugin or extension for Claude memory. Claude isn't a browser app — it runs on Anthropic's servers. Browser extensions don't have a hook into Claude's context. The official system is called MCP (Model Context Protocol), and it's better than any browser plugin would be.

Why there's no browser plugin (and that's fine)

Browser plugins work by intercepting requests in your browser or injecting content into web pages. For something like Grammarly, this works great — it reads what you're typing. But Claude's memory problem is different. The information needs to be on Claude's side of the conversation, accessible server-side so Claude can reason with it — not just injected into your text field.

Anthropic solved this properly with MCP. Instead of a browser plugin, you connect Claude to an external server. When Claude needs to remember something, it calls a tool on that server. When it needs to retrieve something, it searches. The information lives persistently on the server, not in your browser.

What MCP gives you that a plugin couldn't

• Claude actually reads the memory — not just appended text, but a structured search. Claude decides what to retrieve and when.
• Works in every conversation — no need to trigger anything. The connector is always available.
• Scales — you can store thousands of records without bloating your context window. Claude fetches only what's relevant.
• Cross-device — your memory lives on a server, not in one browser on one machine.

How to add Stash as Claude's memory layer

Stash is a hosted MCP server built specifically for this. It gives Claude three tools: add(), find(), and context(). Here's how it works in practice:

Step 1: Sign in with Google

Go to stashlite.com and sign in. You'll get a connector URL unique to your account — something like https://app.stashlite.com/mcp?token=your-token.

Step 2: Add to Claude

In Claude.ai: Settings → Integrations → Add custom integration. Paste your connector URL. Claude now has access to your Stash store.

Step 3: Start using it

That's it. In any conversation:

You: "Save this to Stash: [important information]"
Claude: I've added that to your Stash store. You can retrieve it in any future conversation with find("topic").

— Next conversation —

You: "What do I have saved about [topic]?"
Claude: [Retrieves and presents what you stored]

The context() call: one-line briefing

The most powerful pattern is using Stash's context() tool at the start of sessions. You store standing context — who you are, your current projects, your preferences — and retrieve it with a single command:

context()
→ Returns: role, current projects, working style, tool preferences, open threads
→ Cost: ~150 tokens

This is what replaces the "memory plugin" most people are imagining. Instead of hoping Claude remembers you, you load your context explicitly and cheaply at the start of each session.

Comparison: what actually works

Free tier

Stash's free tier includes 2,500 records and 50 queries per month. For most personal use that's plenty. Pro (£8/month) removes the limits if you're storing more.

← More guides