Bring your clouds.
Connect your own accounts, or talk to us about a managed setup. Keep the flexibility to choose where your files live.
A cloud outage shouldn’t stand between you and an important file. We’re building independent archive copies across providers, with AI-assisted labels and clear rules for what stays. Simple to understand. Yours to control.
Launching soon · No payment or cloud credentials needed
See the real problems behind the product →
Explore how another copy could keep you connected.
AWS • Google Cloud • Microsoft Azure • Dropbox • Google Drive • And more
Planned integrations. Availability will be confirmed at launch.LESS WORRY. MORE POSSIBILITY.
Your files shouldn’t depend on a single provider. Our approach brings your clouds together, so a disruption in one place doesn’t have to mean losing access everywhere.
Connect your own accounts, or talk to us about a managed setup. Keep the flexibility to choose where your files live.
Our planned service creates archive copies across separate providers, with encryption and recovery built into the design.
If a provider is unavailable, recover from another copy. We’re building a simpler path from “where is it?” to “here it is.”
Resilience is our promise in progress. Recovery options, supported providers, and any availability guarantees will be defined in our launch service terms.
MORE THAN A PLACE TO STORE FILES.
We’re building intelligent controls that help you understand your files, protect sensitive information, and keep only what you need.
AI-assisted automation will help you identify files ready for cleanup and define purge policies. Your approved rules—not an AI guess—will determine what can be deleted and when.
Temporary exports Keep for 90 days Eligible for policy review
AI-assisted classification will help identify file types, recognize potentially sensitive content, and apply consistent labels across your archive.
These capabilities are in development. AI classifications will be reviewable and may require correction; labels alone do not grant access. Final retention and deletion behavior will be documented before launch.
WATCH IT. TRY IT. MAKE IT YOURS.
Explore all 18 feature and plan walkthroughs. Watch a 15-second video, then make choices in a guided demo. Every example is fictional; no files are connected, shared, or deleted.
Choose an option to see what changes.
THE 15-SECOND WALKTHROUGH
Silent, captioned concept animation. Play when you’re ready.
Video unavailable. Use the transcript or interactive demo above.
CONTEXT. CONTROL. THE RIGHT MOMENT.
Remember why a file matters, find it from more than one angle, and decide who can access it—and when. Four planned capabilities that make an archive useful long after upload.
Add an archive note alongside a file: what it means, which version was approved, or what the next person needs to know. Notes live separately from the file contents, preserving its original format and existing collaboration workflow.
Maya records “Client approved the blue cover; print proof is in the May folder” beside the final PDF. Months later, a colleague can pick up the reprint without editing the approved artwork or interrupting its shared comments.
Why it matters: decisions survive the handoff.Give a file multiple tags for its project, owner, year, purpose, or review status. Use them to build focused views and identify candidates for an explicitly approved workflow, without creating extra copies.
Leo tags one receipt for the Harbor shop renovation, annual bookkeeping, and the oven warranty. When the oven breaks, he finds the same record through “Warranty” instead of searching three folders.
Why it matters: organize around the question you’re asking.Place selected archive files in a vault with explicitly authorized members. Lock access, review who belongs, and require an authorized unlock. A vault is a separate access boundary; a “Confidential” tag alone does not secure a file.
The Patels want relatives to browse reunion photos while identity documents remain in a vault for two designated adults. The planned vault keeps sensitive archive records out of everyday family browsing.
Why it matters: sharing an archive needn’t mean sharing everything.Plan automatic access to a selected archive file for named recipients at a specific date, time, and time zone. Review, change, or cancel the schedule before release. A release grants approved access; it does not delete the original.
Nora prepares a press kit weeks ahead and schedules access for approved reviewers on premiere morning. Everyone knows the intended release time, while the team can focus on the event.
Why it matters: prepare once and reduce last-minute coordination.All four stories are fictional illustrations, not customer testimonials. These capabilities are planned. Vault controls would cover archive access, not revoke copies already downloaded or change sharing in your source cloud. Scheduled release must respect vault restrictions and holds; delivery timing and retry behavior will be defined before launch.
TRY THE IDEA
Local concept demo with a fictional file. No uploads, real vault, notifications, or background scheduler. Reloading clears your edits. Use made-up information only.
Original PDF · unchanged
Source comments & collaborators · unchanged
Simulation advances this example to the scheduled moment. A locked vault blocks release; editing a form never changes the saved plan until you save it.
ASK YOUR ARCHIVE.
Planned archive intelligence should let you ask practical questions about tags, owners, access, and retention. This preview uses fictional files only; it does not upload, scan, or delete anything.
UNDERSTAND IT BY TRYING IT
Explore two product concepts with fictional files. Nothing is uploaded, analyzed by AI, or deleted. Your choices stay in this page.
Only temporary exports can qualify in this example. Holds always take priority.
| Fictional file | Age | Decision |
|---|
Records the last eight rule changes in this page. Reloading clears this fictional history.
Eligibility means review, not immediate deletion. This demo does not define a legal retention period or operate on real files.
Select a fictional sample to explore how classification could support organization and access review.
No review recorded for this sample.
Preset examples, not an AI analysis. Labels are suggestions: review them before applying access or retention policies.
PERSONAL FILES. PROFESSIONAL PEACE OF MIND.
The first photos. The final draft. The documents you hope you’ll never need. Give your digital life a little more protection.
Client projects, company records, years of hard work. Plan for continuity without making cloud storage another full-time job.
A USEFUL FIRST STEP
Build a simple plan, estimate copy capacity, and explore costs using your own assumptions. No signup needed. No details are sent.
Save settings on this device, copy a brief, or share a scenario. Names, email addresses, and inquiry text are never included.
Settings are not saved automatically. Reset changes this page only; Forget removes the saved settings.
A LITTLE PLANNING GOES A LONG WAY.
Choose a starting point. We’ll help you find the right fit.
Proposed launch pricingPERSONAL
A simple start for the files that mean the most.
SMALL BUSINESS
A stronger foundation for your business files.
CUSTOM
More files, more clouds, or a different kind of challenge.
Estimates for our upcoming launch, not an offer to purchase. Features, storage allowances, and final prices are still being defined. Cloud-provider storage and transfer charges may apply separately. No payment is collected.
REAL PROBLEMS. A CLEARER PLAN.
Historical Reddit discussions reveal worries that storage space alone doesn’t solve. Here’s how they shape Cloud Archival.
Community reports are individual experiences, not verified customer outcomes or endorsements. Our service is prelaunch; each response below is marked as a planned capability, a design requirement, or a tool you can try today.
During the June 12 Google Cloud outage, Reddit users described disruption to their users and customers. Google’s incident report lists Cloud Storage among affected services.
Keep an independent copy on another provider, such as Amazon S3, so there can be another way to retrieve an already archived file.
A second copy needs a working way to sign in, find the file, and unlock it. Identity and key services were also affected. Our recovery design must avoid relying on the unavailable provider for those steps. This has not yet been validated in a working service.
Reddit discussion · June 12, 2025Google’s official incident report
Try the GCS outage illustration →A Google Photos user described nearly losing vacation photos while trying to manage cloud and device copies. The discussion exposes how confusing synced deletion can feel.
Independent archives and explicit deletion policies are the intended response. Our policy demo shows what a rule would affect before anything is eligible for removal.
Two locations are not two safe copies if the same deletion reaches both. Source deletions must be handled separately from archive retention. The service’s exact behavior and recovery windows still need implementation and testing.
Reddit discussion · June 19, 2023
Try the deletion policy demo →A person helping an elderly relative described family photos behind a forgotten password, with recovery codes going to an old phone number.
Separate copies need a separately usable access path. Cloud Archival’s recovery planning must account for account access as well as where files are stored.
If every recovery path depends on the same locked account, adding another cloud may not help. Independent authentication and key access remain design requirements, not shipped features. We cannot unlock a provider account or recover a file that was never copied.
Reddit discussion · April 21, 2026
Read our recovery limits →A DataHoarder discussion raised the extra steps and costs involved in retrieving a deep archive. AWS documents different retrieval options and waiting times by storage class.
Our cost worksheet and transfer estimator help you explore your own assumptions before choosing an archive setup. They already work here, without an account.
Transfer time starts after the file is ready to download. Retrieval queues, restore requests, temporary storage, and provider charges can add time or cost. The tools are estimates, not a measured restore or a recovery guarantee.
Reddit discussion · February 11, 2021AWS archive retrieval documentation
Estimate a recovery transfer →Simple on purpose. AI is planned to help label files and flag potentially sensitive content. You review its suggestions and control deletion rules. Outage recovery depends on usable copies and access—not an AI prediction. An internet outage would still require a previously downloaded local copy; offline access is not currently a feature.
CLARITY BEFORE COMMITMENT
Trust starts with knowing what exists, what’s planned, and what still needs to be proven.
Our design starts with your cloud accounts and your policies. Connection permissions, export options, and account-disconnection behavior will be documented before launch.
Encryption and limited access are design priorities. We are not claiming a completed security audit or compliance certification. Technical controls will be published before accepting files.
Independent copies are intended to reduce reliance on one provider. Storage tiers, working credentials, and verified copies affect recovery. No uptime or recovery-time SLA is available yet.
A FEW THINGS YOU MIGHT BE WONDERING.
Planned archive notes will be separate metadata alongside the original file. They will not edit the document or alter its source-cloud comments and collaborators. Note visibility will follow archive permissions, with final behavior documented before launch.
Yes, that is the planned design: combine project, owner, year, purpose, and status tags on one file. Tags help filtering and policy review, but do not grant access, bypass a vault, or authorize deletion by themselves.
No. The planned vault controls access within Cloud Archival. It cannot revoke downloaded copies or replace permissions at your source provider. Vault locking is an access control concept, not a claim of immutable storage or certified regulatory retention.
The planned service will evaluate the scheduled date, time zone, named recipients, permissions, and any vault restriction or hold. A blocked release must stay blocked until an authorized person resolves it. You will be able to review or cancel pending schedules. This website only simulates the decision; it cannot release files or notify recipients.
That’s the plan. We’re designing Cloud Archival to work with your own cloud accounts. Tell us which providers you use so we can discuss launch availability and your setup.
Our planned approach keeps copies with separate providers, so you can recover from an available copy. Actual recovery times will depend on the storage tier, provider availability, and your plan. Service commitments will be published before launch.
Encryption, limited access, and independent copies are core design priorities. Technical security details and data-handling terms will be published before we accept files. Please don’t send passwords, account keys, or sensitive documents in an inquiry.
We’re preparing for launch and welcoming early-access inquiries. File uploads and customer login are not available yet. Joining the waitlist does not require payment or commit you to a subscription.
The prices shown are proposed service estimates. Storage allowances and how provider storage, retrieval, and transfer charges will be billed have not been finalized. We’ll confirm all costs with you before signup.
Our planned design requires an explicit approved purge policy. AI can suggest labels or cleanup candidates; it will not replace your approval rules. The sandbox uses fictional files and deletes nothing.
No. The demonstrations use preset examples and browser calculations. There is no file uploader or AI service behind them. Planner entries stay in the page unless you choose to include them in an email you send.
Review the email draft and send it to mahesh@mahesh.fyi. We will respond to discuss your needs and launch availability. Opening or downloading a draft does not send it or reserve a place.
No. Planned labels will be reviewable suggestions. Permissions must follow approved access policies. AI accuracy, processing locations, and data-handling details will be documented before file processing begins.
No matching questions. Try a shorter term, or ask us directly.
YOUR FILES HAVE A FUTURE.
Tell us what matters to you.
We’ll help you plan what comes next.