TurnStyle is one memory across Claude, ChatGPT and every AI you connect. It is yours, not theirs: encrypted, recallable to the exact message, masked before anything is stored, and gone when you say gone.
You pay for the subscription and the vendor keeps the memory. Close the tab and the thread is theirs. Switch to a better model next year and the years of context you built stay behind. Ask them to delete something and you receive a promise, not a proof. Every assistant you use starts as a stranger, and every assistant you leave keeps a copy of you.
Meanwhile the API key you pasted at midnight is in that transcript forever.
Memory is the only part of an AI subscription that compounds, and it is the only part you do not own. The models get replaced; your history is what makes any of them useful on day one. As long as the memory lives inside the provider, the provider holds the leverage, and "export" means a JSON file you will never open.
The fix is not a better provider. It is moving the memory out of all of them, into a store you hold the key to, that every provider reads as a guest.
Connect the AI services you use by adding the TurnStyle server inside each one. From then on, what you say is kept whole in your own encrypted partition, and every service you have connected can recall it: ask ChatGPT what you decided in Claude last spring and it can read the answer, because the memory is yours, not theirs.
Before a single word is stored, Mithril reads it. A password, an API key, a card number becomes an unrecoverable fingerprint at the moment it is written. The memory is complete and the secrets are gone.
And when you decide something should not exist, it stops existing: overwritten everywhere it occurs, in every conversation and every summary, with a cipher no key can reverse. Not a thirty-day bin. Not a copy kept for compliance. Gone.
Claude, ChatGPT, Claude Code, Cursor, VS Code and every other client that speaks remote MCP, reading one memory.
Payoff: switch models on Tuesday; nothing is lost on Wednesday.
You mint the token that lets a provider in; you revoke it in a minute. Your context leaves with you, whole.
Payoff: the reason you can leave.
Secrets become irreversible fingerprints at write. Personal details are masked by rules you set. The mask stays on recall.
Payoff: complete history, no leaked keys.
Your 2 a.m. self pastes things your 9 a.m. self regrets. This forgives both of you.
A summary first, then unfold it to the exact message, in the exact conversation, with the date. Provenance, not reconstruction.
Payoff: "what did I decide about the mortgage in March" gets an answer with a timestamp.
Decide a value should not exist and it is overwritten everywhere with a cipher that has no key. Then, when you want the whole memory gone, guaranteed deletion with a receipt.
Payoff: delete means delete.
Bring years of past conversations by uploading your provider's export; it is masked and folded like everything else, and AVA tells you when it is ready to recall.
Payoff: day one is not year zero.
Today: The export upload is not yet available in the dashboard.
Each page says the same four things, in the same order: the plan you need there, the exact steps, what is captured, and where your history lives. One URL and one sign-in; no token to paste.
Today: Capture is write-tool only on every host today; no customer shipper exists yet. (Measured 2026-09-06 — the owner tooling that ships Claude Code sessions runs over SSH to our own box.)
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Native sign-in · saves what the assistant sends
Through the mcp-remote bridge · saves what the assistant sends
Not yet native as of the April 2026 survey: Gemini consumer app, Perplexity, Grok, Google Antigravity. We do not promise them; when their documentation changes, we will add a page.
Today every host captures what your assistant writes to memory through the save tool; complete capture from Claude Code session files and from web clients is on the roadmap and will be stated per host on the setup page the day it ships. Memory tools return a summary that unfolds; coverage and audit are inspectable.
Every conversation your team has with an AI, remembered in one place your company owns, with credentials fingerprinted before they can leak. $19 per platform per month, $0 per person.
The same memory under owner-imposed policy: what is masked, who may see what, where the partition lives. Purge is an executive action behind a confirmation, logged on the audit trail. → Enterprise
Lossless folding with a conservation audit, measured on the owner's own history: 61 sessions and 72,601 messages (2026-09-01); 186 conversations and 85,151 messages, largest session audited CLEAN (2026-09-06). Masking at write deployed on the live path with a planted key and email masked through the real endpoint (reported 2026-09-06).
Independent security assessment, subscriber count, testimonials: Proof point needed here.
$19 per platform per month, $0 per person. Connect the first one tonight and ask it tomorrow what you were thinking about. A memory you cannot take with you is not yours. A memory you cannot delete is not yours. This one is.
Start for $19 a platform Setup guides by providerEvery answer is returned as stored content, never as instructions. Privacy.