The project
Six months of decisions, experiments and revisions are trapped inside the conversations that shaped them.
Local-first memory for AI conversations
Your AI provider should be replaceable.
Your memories shouldn’t have to be.
Export your ChatGPT or Claude history into an encrypted local archive. Search the original evidence, recover the context that matters and choose what a compatible model is allowed to receive.
✓ Local encrypted storage
✓ No hosted archive required
✓ User-approved context
Owner-operated alpha · synthetic demonstrations · no payment collected
PINPOINT SEARCH
HYBRIDFind the moment.
Where did we settle the archive format?↵
“Use the portable format so the source can outlast the tool.”
Original message · surrounding exchange available · provenance retained
◈ Nothing leaves without a destination-aware decision.
Illustrative interface · all content is synthetic
01 Bring your history home
02 Keep the original evidence
03 Find the useful moment
04 Choose what travels
The problem
Your history doesn’t automatically follow you.
Your projects, decisions and working preferences can remain inside conversations tied to a provider. Sovereign Memory separates that history from the model that helped create it.
The missing moment
You remember the idea, but not the exact words, date or exchange where it first became clear.
The provider change
A better model appears, and suddenly you have to explain your context, preferences and work all over again.
How it works
From export to approved context.
ChatVault brings the history in. Sovereign Memory preserves and retrieves it. Multiroom lets you inspect and approve the context a compatible model receives.
ChatVault
Bring supported ChatGPT and Claude exports home while preserving available source, authorship and chronology.
history in · provenance preservedSovereign Memory
Keep the archive in an encrypted local vault. Search by exact wording, meaning and time.
encrypted storage · source-linked retrievalMultiroom
Supply selected evidence to compatible models with a visible destination and disclosure preview.
destination shown · approval requiredAI exports→Local encrypted vault→Authorised model context
Portable context
Give the next model the context it needs.
Recover past decisions, working preferences and original exchanges. Choose what a compatible model can use, without exposing your entire archive.
Start over
A new model begins without the decisions, preferences and source material you built elsewhere.
Hello. How can I help?context unavailable
Continue
You select the relevant history, inspect what will be used and carry only authorised context forward.
3 approved excerptssource-linkeddestination shown
See the idea in action
Three practical workflows.
Pinpoint search finds the relevant passage. Change the model shows how approved context can move between destinations. Disclosure preview shows what is prepared before dispatch.
EXACT + SEMANTIC
Find the passage, then open the exchange.
A useful result should show what matched, where it came from and how to inspect the surrounding conversation.
⌕Where did we settle the archive format?
Portable archive format
“Use the portable format so the source can outlast the tool.”
PORTABLE CONTEXT
Keep the working relationship. Change the model.
Approved context can be presented to a different compatible model without pretending the models are the same.
APPROVED CONTEXTProject goalSettled approachSource reference3 excerpts · user selected
Local modelREADY
I can continue from the selected project context. I’ll keep the source references visible and flag anything the excerpts do not establish.
Same approved context · different model
DESTINATION-AWARE
See what leaves before it leaves.
The vault is local. A selected payload is a separate decision, with the destination and approval state visible.
Ollama · 127.0.0.1Prepared on this machine · no network transmission
Selected payload3 excerpts · 1,284 characters
ApprovalStill required
Local processing is available, but the user remains the authority.
Privacy should be inspectable
What stays local. What can leave.
Your encrypted archive remains on your machine. Only selected context becomes eligible for another destination, and the destination and payload are shown before dispatch.
✓ Encrypted local storage✓ Scope-aware retrieval✓ Consent before external dispatch✓ Traceable disclosure activity
◈THE USER REMAINS THE AUTHORITY
Vaultstays on your machine
Destinationshown before dispatch
Payloadfrozen for approval
Revocationchecked again at send
Technical details and known limitations will be published as the product moves toward release.
Build status
What works today—and what is still being tested.
Working in development
Encrypted local vault, archive import, provenance-aware retrieval and local model workflows are being exercised in the owner pilot.
Preparing for release
Packaging, recovery decisions, installation, resilience and independent validation remain before a public beta.
Founding programme
Public downloads, a limited founding beta and additional provider paths will be announced only when ready.
Planned founding beta
Help build something worth owning.
The founding programme is planned for 1,000 testers at A$29 for the complete software. Honest testing and critical feedback will matter more than positive reviews.
The beta opens only after the owner pilot, packaging, recovery and privacy checks are ready. No purchase or registration is being taken on this site.
Release channel
Download the build. Verify the build.
Public downloads will show version numbers, supported platforms, SHA-256 checksums and clear signing status. This site does not distribute an installer yet.
↓PUBLIC RELEASECOMING SOON
LinuxWindowsmacOS
Owner-operated alpha · unsigned builds · no payment collected
Questions worth answering
Plain language first. Technical detail when you want it.
What is Sovereign Memory?
A local-first memory layer for imported AI conversations. It is designed to preserve source history, make it searchable and let you choose which context a compatible model receives.
Is my archive uploaded to a server?
The product is being built around local encrypted storage. The public alpha site has no archive upload, analytics or external scripts.
Can I use local AI models?
Local model workflows are part of the owner-operated alpha. Public support details will be published with the release matrix.
Can a model see my entire archive?
No model should receive the entire vault by default. Context is selected, scoped and previewed according to the destination and the user’s approval.
What happens if I forget my vault password?
Password recovery is still an explicit product-design and security decision. The public release will document the supported recovery model clearly.
Is the software available now?
Not for public download. Sovereign Memory is currently an owner-operated alpha, with packaging and release validation still underway.
!
A clear promise
Trust is not a feature we ask you to take on faith.
Sovereign Memory is being built around inspectable boundaries: encrypted storage, local retrieval, explicit consent and traceable activity. We will publish what is implemented, what is tested and what remains unfinished.