Encrypted · tenant-isolated · you control access

Capture once,
retrieve anywhere.

Memory Vault is a private, encrypted store for the memory your AI assistants build up — notes, decisions, project context. Connect your tools, see every memory, and stay in control across every model you use.

No credit card. Your vault stays encrypted and yours.

Encrypted at rest

Per-tenant data keys, KMS-managed. Raw content and tokens are field-encrypted.

Tenant-isolated

Every read and write is scoped to your vault — enforced in the app and the database.

Fully auditable

Every key use, ingest, and deletion is logged. Nothing happens to your memory in the dark.

You control access

Scoped API keys per tool. Revoke any connector instantly — it stops syncing the moment you do.

How it works

Your AI memory, in one place you own

Connect the tools you already use. Memory syncs into a single encrypted vault you can browse, search, and control from any device.

1

Connect a tool

Create a scoped key and point your agent at the vault — mv connect. Works with MCP servers, SDKs, and the browser extension.

2

Memory syncs automatically

As your assistants work, notes and decisions flow in — each one tagged with where it came from and which project it belongs to.

3

Browse, search & control

Search across every tool, edit or pin what matters, export anytime, and revoke any connector in one click.


Features

Built for people who live in their tools

Practical controls for technical work — not a magic black box.

Hybrid search

Keyword, semantic, or both. Find the right memory across every connected tool in milliseconds.

Source provenance

Every memory shows the tool and connector it came from. You always know where a fact originated.

Export anytime

Download your entire vault as JSON — every memory, tag, and metadata record. No lock-in.

Scoped keys & revocation

Issue a key per connector with just the scopes it needs. Revoke instantly — syncing stops at once.

Audit log

A read-only record of every action — key use, ingest, deletion — optimised for scanning.

Projects & tags

Memory is organised by project and topic, so the right context surfaces without bleeding across work.

Use cases

Your context, wherever the question finds you

Most of your project knowledge lives inside AI conversations on one machine. Memory Vault syncs it to a vault you can reach from any device — so you're never stranded without it.

For freelancers & consultants

Answer a client from your phone — accurately

You're away from your laptop when a client emails for a status update. Open Memory Vault on your phone, search the project, and reply with the real decisions and current state — pulled straight from the memory your tools captured while you worked.

  • Search every decision, note, and summary across tools
  • Works on any device — phone, tablet, a borrowed laptop
  • Provenance on every memory, so you can trust what you quote

Developers

Carry architecture decisions, gotchas, and conventions across Claude Code, Cursor, and every repo. Stop re-explaining context each session.

Founders

Keep product decisions, customer notes, and strategy in one searchable place — across every tool and model you bounce between.

Consultants

Each client's context stays scoped and searchable, so you can switch engagements and still answer precisely.

Researchers

Sources, findings, and threads from long investigations stay linked and retrievable — not buried in old chat logs.

Works with your stack
Claude CodeCursorChatGPTBrowser extensionPython & TS SDKMCP servers
Pricing

Start free. Scale when you do.

Priced on what you store and sync — never per seat.

Hobby
For a personal vault and a single tool.
$0/ forever
  • 1 connected tool
  • Up to 500 memories
  • Keyword search & export
Start free
Builder
For teams and heavy programmatic use.
$89/ month
  • Everything in Pro
  • Higher API rate limits
  • Shared team vaults & SSO
  • Priority support
Contact sales

Own your AI's memory.

Set up a vault in minutes. Connect a tool, watch the first memory sync, and never lose context between models again.