One memory. Every AI.

The portable, searchable memory layer that follows you across ChatGPT, Claude, Gemini, Perplexity, and Grok — stored 100% locally, so it's yours even when you switch vendors.

In 30 seconds

Every major AI tool has its own memory — ChatGPT capped around 1,500–1,750 words, Claude scoped per project, Gemini tier-gated, Grok barely there — and none of them talk to each other. Memdex is a Chrome extension that gives you one portable memory layer instead: every conversation, captured locally and encrypted, surfaced in whichever AI you're currently typing into.

  • One memory across all your AI tools — not five siloed, capped, and unreliable ones.
  • Kept 100% locally in your browser — no TEE, no cloud, no Memdex server ever sees the content.
  • Surfaces relevant past chats as you type, regardless of which tool they originally happened in.
  • Survives the next vendor switch. Claude-native memory dies the day you switch to whatever's next; Memdex doesn't.

Native AI memory is built for one tool, not your workflow

Native AI memory is three things at once: capped, opaque, and vendor-locked. When ChatGPT's memory wiped in February 2025, nobody had a copy. When you leave Claude, its project memory stays behind. A portable memory layer is the only design that survives the fact that you use five different AIs and probably will use seven next year.

SILOED

Every AI has its own memory — none of them know each other

What ChatGPT remembers about your startup doesn't carry to Claude. What you told Claude about your codebase doesn't carry to Gemini. You repeat the same paragraph of context three times a week.

CAPPED

The capacity is smaller than you think

ChatGPT's memory caps around 1,500–1,750 words per user. Claude's project memory is scoped per project. Gemini's is Advanced-tier only. You will hit the ceiling and you will be asked to delete something you care about to make room.

UNRELIABLE

When the vendor flinches, you lose everything

In February 2025, a GPT-4o bug wiped persistent memory for many users. OpenAI fixed it, but the wiped context never came back. Native memory is at the mercy of the vendor — yours doesn't have to be.

PORTABLE

What you actually need is memory that travels

You want one memory that works the same in Claude as in ChatGPT, that you can inspect and export, and that stays with you if you switch tools next month. That's a portability problem, not a feature problem.

How Memdex gives you one portable memory

Capture everything once. Surface it everywhere. Keep the whole thing on your device.

01UNIFIED CAPTURE

One local index of everything you've said to any AI

Once installed, Memdex captures every conversation on ChatGPT, Claude, Gemini (native), Perplexity, and Grok (shortcuts) into a single encrypted local index. There are no five memories — there is one.

02INLINE DETECTION

Relevant context surfaces in whichever AI you're using

When you start typing in any supported AI, Memdex detects keywords from your unified history and highlights matches. You don't search — it suggests.

03ONE-CLICK INJECT

Pull exactly the right thread into the current chat

Click the match to attach the full conversation as context to your current prompt. The AI you're in now reads it as a proper dialogue. Your project never starts from zero again.

How Memdex compares to other 'AI memory' tools

Most memory products either live in the cloud (fast, but someone else holds your data) or target developers (capable, but not built for a power user typing in Claude). Memdex is the consumer local option.

CapabilityMemdexMemoryPluginAI Context VaultMem0
Auto-captures conversations as you have them
Works inside ChatGPT, Claude, and Gemini natively
Stores memory 100% locally (no cloud, no TEE)
Surfaces relevant memory as you type
Built for end users, not developers
One-click inject full past conversation

MemoryPlugin uses cloud storage with optional TEE. AI Context Vault stores locally but relies on manual context tagging. Mem0 is a developer memory API, not a consumer extension. All comparisons accurate as of April 2026.

Frequently asked

Native memories are siloed (each tool keeps its own), capped (ChatGPT around 1,500–1,750 words, Claude per-project, Gemini tier-gated), and vendor-locked (they don't leave the tool). Memdex is one portable memory that works across all of them, with no hard cap and no vendor lock-in, stored on your device.

Stop explaining yourself five times.

One memory, every AI — ChatGPT, Claude, Gemini, and more | Memdex