More than AI memory

Not just facts about you.

Actual conversations and files from past chats.Your AI work continues across sessions - and platforms.

Sync it. Connect it. Your AI can go back and get what it needs.

Start FreeWatch the demo

Free saves your last 7 days. No card required.

See it in action

Three situations you already recognize. Each one ends with the next session knowing what the last one did - in the same AI, or a different one.

“Another agent had some good advice I thought about not apologizing for asking for money on the end. Can you go search for that please?”

No platform, no date, no name. It came back with the exact paragraph - from Perplexity, inside Claude - then checked it against the draft that was open at the time.

“You just searched 3GB of my personal AI conversation history and found the piece of a conversation I had on another platform. I’m in shock.”

A real session

A real Claude session: the user asks for advice they half-remember from another AI, and Claude returns the exact paragraph from a Perplexity session weeks earlier.

Continue. Find. Bring it in.

Everything else on this page is detail. If these three work, you stop being your AI's memory.

Continue

Open a new chat. Pick up where the last one stopped.

Not a summary. Not copy/paste. All of last session’s details are immediately available to your AI - the way you’ve always wanted.

When it tells the new session what you decided, it points to the message where you decided it - and reads that message before it says anything.

What the new session actually receives
DecisionKilled usage-based pricingmsg 39
DecisionKept seat + usage capmsg 41
OpenLaunch date not setmsg 52
VerbatimLast 5 messages, unabridgedmsgs 48-52

The shape of a real briefing. Every line is pinned to a message; nothing is quotable until it has been read.

Find

Ask for the thing you barely remember.

“What did we decide about pricing in March?” “Where is that line about not apologizing for money?” One search across every AI you have connected, and it comes back with the exact words and where they came from.

It searches the transcript, not a summary of it. That is why it can correct itself: in one live session the agent stated a detail with full confidence, then reversed it after the vault surfaced the actual decision from three months earlier.

A real Claude session: asked whether thread ids are now immutable UUIDs, the agent searches the transcript history and corrects the answer it gave a moment earlier.
A real session, underline ours. It searched the transcripts, found the March decision, and corrected itself - unprompted.

Bring it in

Pull the file, the paragraph, or the decision into the chat you are in now.

The file it made you last Tuesday is still there. So is the screenshot you gave it three weeks ago. Ask your AI for it and it lands in the current conversation. Nothing re-explained, nothing shared twice.

File sync is optional and per account. Turn it on and your vault saves the file itself, not a link to it, alongside the message it came from.

Files from one thread
pricing_model.mdAI wrote
competitor_teardown.pdfyou shared
kickoff_notes.docxyou shared
onboarding_flow.pngyou shared

Files you shared and files the AI wrote, saved with the message they came from.

Read the story: Same thread. Different brains.

Not memory. Not an exporter.

Built-in memory

Your AI’s built-in memory is a few pages of notes about you - your name, your job, a paragraph on what you generally work on. Nothing about your project, your decisions, or anything you have actually talked about. And it stays inside one app.

Chat exporters

A chat exporter gives you a folder of files you will never open. It protects the past and does nothing for the next session - the AI still starts from zero.

ContinueVault saves the work itself - the conversations, the decisions, the files - and your AI brings it back inside whatever session you are in, on any AI you use. It is not tied to one provider, so you get more out of the AI subscriptions you already pay for.

Setup: two minutes.Then it runs in the background.

This is not a new chatbot. You keep using ChatGPT, Claude, and the rest. ContinueVault sits beside them, saves the work, and hands it back when a session starts over.

  1. Step 1

    Create an account

    Free tier, no card. Upgrade later if you want your full history saved.

  2. Step 2

    Install the Chrome extension

    It only runs on the AI sites you approve. Tokens stay in the browser.

  3. Step 3

    Connect the accounts you want synced

    Pick project accounts. Leave personal chats out. File sync is a per-account switch, off until you turn it on. CLI tools default to conversation-only.

  4. Step 4

    Open a new session and say “continue”

    Use MCP in Claude, Grok, Perplexity, or Le Chat. Use the ContinueVault GPT in ChatGPT. No new chat UI.

Then try the real test: open a new chat and ask about something you barely remember from last week. That first hit is the whole product.

Start with the last week. Save everything when you’re ready.

Free is the whole product, saving your last 7 days of chats and files. Pro is the same product with your full conversation history saved.

Free
$0/mo
The whole product. Last 7 days.
  • Save and search everything you sync
  • Continue Thread - pick up where you left off, within your 7-day window
  • Export anytime
Get Started
Recommended
Pro
$19/mo
Founding price, for the life of your subscription
The same product. Full history, nothing expires.
  • Everything in Free
  • Full history - chats and files stay and can be referenced
  • Continue Thread across everything you’ve ever saved
Subscribe

Pro saves all of it.

Sync everything you have ever done on the accounts you connect. Decisions, dead ends, and open items come forward - not a bigger pile of old chats.

ChatGPT, Marchpricing model debate…
Claude, Aprilauth flow rework…
Perplexity, Mayvendor comparison…
decision kept
dead end marked
still open

Free covers the last week. Pro saves everything you have ever synced.

Your AI work stays yours.

ContinueVault stores your conversations in a per-user encrypted vault. We don't use them for training, and you can export or delete everything on any tier.

The important trust difference is architectural: your vault is decrypted only by a server process actively handling your authenticated requests. When you go idle, that process shuts down - taking the keys with it, typically within minutes. There are no scheduled jobs reading your conversations between sessions, and no support access without your explicit, time-limited consent.

ContinueVault saves the conversation around your AI work, not your code by default. CLI sources sync prompts, decisions, and reasoning by default - not file contents, diffs, command outputs, or tool payloads. Full session sync is an explicit opt-in.

Read the security architecture

What works today.

Everything you connect is saved and searchable. File sync is optional; where a platform makes files available it saves the files too, and where it does not yet, the row says so. Picking the thread back up in-place works in the clients marked below. We would rather show the gap than hide it.

Claude
Save
Yes, chats and files
Continue there
Yes, via MCP
ChatGPT
Save
Yes, chats and files
Continue there
Via the ContinueVault GPT
Grok
Save
Yes, chats; files in progress
Continue there
Yes, via MCP
Perplexity
Save
Yes, chats
Continue there
Yes, via MCP
Le Chat
Save
Yes, chats
Continue there
Yes, via MCP
Gemini
Save
Yes, chats
Continue there
Not yet; continue from a row above
DeepSeek
Save
Yes, chats
Continue there
Not yet; continue from a row above
Claude Code / Codex CLI
Save
Conversation-only by default
Continue there
Continue in MCP clients or the GPT

Compatible tools can also connect with your vault URL. Full session sync on CLI is an explicit opt-in - source trees stay in git. No native mobile-app sync yet: web chats via Chrome, plus Claude Code and Codex CLI locally.

Where it came from

It started as a question to ChatGPT.

Jeremy Waldrop, founder. Thirty-five years of embedded systems and firmware that had to work.

In November 2025 I was using ChatGPT to iterate on code that drew schematics. Draw, look, tweak, draw again. An hour in, the context was full and the session was tired.

So I asked it the only question that mattered: what do you need to tell the next session so it can pick up where you left off and do the job as well as you are doing it now? It wrote notes to its successor. The next session read them. We did that again and again until it had a whole system for it - lessons learned, decisions, dead ends, what to do next.

ContinueVault is that question, asked automatically, every time. I built it for myself first. It became a product because it turned out other people were tired of losing their work too.

The rest of the story

Stop being your AI’s memory.

When a session hits the wall, the next one - on any AI you use - picks up right where you left off.

Start Free