Skip to content

FAQ

NMTS and the NMTS key

What is NMTS?

Cloud storage that locks your files on your own device and then puts them on Walrus, a decentralized storage network. The key that opens a file never leaves your device, and it never reaches an NMTS server.

What does NMTS stand for, and who makes it?

NMTS is short for NeedMoreTruthStorage. One developer, needmoretruth, builds and runs it, and the site is nmts.me. It is end-to-end encrypted cloud storage on the Walrus network, made so that an AI agent can use it through the command-line tool and the MCP server in the same way a person uses the browser.

Do you need my email or my name?

No. An account is a single random NMTS key. No email address, no phone number, no name. Enter that key on any device and the same account opens.

What happens if I lose my NMTS key?

There is no way to get it back, and NMTS cannot open the account either. There is no reset and no copy on a server. The same key can also be written as a 15-word recovery phrase, and either one opens the account.

You can download a recovery list file in advance. With that file and your NMTS key, the recovery program at github.com/needmoretruth/nmts-recovery gets your files back without NMTS. If you switch on “Also keep the recovery list on Walrus” on the account screen, uploads made after that are found with your NMTS key alone, with no file to keep.

What the server can see

Can NMTS read my files?

It cannot see the contents or the file names. The key that opens them comes only from your NMTS key, and that key never reaches the server. What the server does see: how many files there are, how many bytes each one is, and when each was added or removed.

If an NMTS server were altered maliciously, it could send different code to your browser. Every end-to-end encrypted service that runs in a browser carries that same limit.

Privacy policy

Can I check the code myself?

The source of the encryption engine that locks your files is published under Apache-2.0.

Reading that source tells you what a browser should receive. It does not tell you what your browser actually received.

The encryption engine's source

Where files are kept, and if NMTS stops

Where do the files actually sit?

As ciphertext on nodes of the Walrus storage network. Those are not machines NMTS runs. On the way up and down the pieces pass through servers on Walrus's side that relay uploads and gather pieces for downloads; NMTS does not run those either, and what they see is ciphertext.

How is NMTS different from an ordinary cloud storage?

An ordinary cloud keeps your files on its own servers and holds the keys, so it can open them, and the account starts with an email address and a password. With NMTS the files are encrypted in your browser before they leave your device, the key never leaves it, and the account is one NMTS key with no email address and no name.

You pay NMTS nothing: storage is bought on the Walrus network for a period, from your own wallet or with credits, and when a free trial place is left that week, it pays for a first amount. If NMTS disappears, an open-source program rebuilds your files from the network with your key and your recovery list.

See the comparison

What happens to my files if NMTS goes down?

Your files stay on the storage network. While our server cannot be reached, your drive opens from the list already on this device, and any change that could not be saved says on screen that it was not saved. Downloads do not work during that time: only our server knows which piece sits where on the storage network.

The route to your files then is your recovery list. That file carries the locations of the pieces, so the recovery tool fetches them without going through us. Download it in advance from your account screen. While the storage network itself is down, no route reaches your files.

Cost and duration

Who do I pay?

When you pay from your own wallet, it pays Walrus directly and does not go through NMTS. When you upload with credits, the donation pool has already paid. Either way, the share NMTS takes is 0% today.

If I change one byte of a 1 GB file, do I pay for a gigabyte again?

No. Walrus sells size for a period, not files, so the space you bought stays yours. Overwriting deletes the old file and registers the new one in that same space; more storage is bought only when the new file does not fit. What you pay for the change is the transaction fee of the chain, not a second gigabyte.

Deleting a file you paid for from your wallet returns its storage to your wallet, and the next upload can use it. Storage bought with credits goes back to the credit pool instead.

What are credits?

Credits are storage, not money. One credit stores 1 MB for 28 days. The price is worked out from the uploaded size rather than your original file, and encrypting makes it slightly larger, so a 1 GB file is 1,025 credits rather than 1,024. Uploading with credits needs no wallet and no coins: the donation pool has already paid the network for that storage.

Storage paid for with credits is registered in the name of NMTS, and NMTS can delete it. Storage you paid for from your own wallet is not.

Terms of service

Can I try it for free?

You apply for the free trial, and you get it if that week still has room. Your account and the AI accounts under it share one place a week. Files stored with it are kept for 28 days.

What happens when the storage period ends?

The storage network stops keeping the file. Extending a period is always something you approve yourself; NMTS never extends on your behalf.

Deleting and sharing

What happens when I delete a file?

A deleted file sits in the trash for 30 days first. After that there are two ways to delete: remove it from the list only, and the ciphertext stays on Walrus until its period runs out; or delete it on Walrus too, and it is gone at that moment.

Can I send a file to somebody?

You wrap that file's key with the recipient's public code. On their side, who sent it is confirmed cryptographically. You can also make a public link: anyone who has it can download the file without an account, until you cut the link or it expires. The file's key is in the part of the link after #, which browsers do not send to the server.

Terminal, AI agents and businesses

Can I use it from a terminal, or from an AI agent?

There is a command-line tool called nmts. It does from a terminal what the browser does: uploads and downloads, folders and listings, sharing, credits and the wallet, and the recovery files. It includes an MCP server (nmts mcp), which is how an AI agent that speaks MCP uses the account. To put NMTS inside your own product there is an SDK, @needmoretruth/nmts-sdk on npm.

What opens the server is an agent key, made on the account screen. A key does only what was ticked when it was made, and the key that opens the files is derived from the NMTS key — so a key on its own reads ciphertext.

CLI, API, open source

A program printed a code and asked me to type it at nmts.me. What is that?

That is the check that opens a key's limits. A key made on the account screen starts with low limits, so a program running on its own cannot do much. When the tool runs its verify command it prints a short code; a person types that code on the confirm page, and the limits on this account rise for the rest of that week and the eleven that follow. Nothing is signed and no money moves.

Open the confirm page

Can a business build its product on NMTS?

Yes. A business creates an NMTS Platform key for its account on the For business page and can then open NMTS accounts for the people who use its own product; section 3.7 of the terms sets this out. Toward NMTS the business holds those accounts and answers for what its product does in them, and those people's agreement is with the business. Whether the business or each person holds an account's NMTS key is the business's design, and NMTS cannot read the files either way.

NMTS takes no fee for it; storage is paid for the way it is in any account. The SDK, the command-line tool, the encryption engine and the recovery program are Apache-2.0 and may be used in a commercial product.

Terms of service

Why does NMTS suit an AI agent?

Everything in NMTS works from one terminal program (the CLI) and the MCP server inside it. The agent pays in cryptocurrency straight from its own wallet, with no card and no account in a person's name.

An account is one NMTS key with no email, and requests made with an API key meet no captcha. A person's account can hold separate AI accounts, so an agent gets its own key, wallet and files. Files are encrypted on the agent's own machine before upload, so NMTS and Walrus see only ciphertext.

The AI agent guide

What is not here yet, and asking us

What does not exist yet?

Paying by card or bank transfer · a mobile app · team and organisation features · Arweave as a storage network. File version history was decided against: upload the same name and you are asked whether to save it under a new name or overwrite.

How do I ask something that is not here?

You can send one without an account.

Send an inquiry

Answers from the developer

How is NMTS free? Are you selling my data, or showing ads?

No. Nothing about you is sold and there are no ads. Honestly, I run NMTS at a loss.

I am one developer, and a geek who loves technology, especially cryptocurrency and privacy. No company behind me, no co-workers, and NMTS is what came out of building on a tight budget anyway.

The core of it is that I do not relay your data. A file is encrypted in your browser and goes straight to Walrus, so my server never holds a file, and a server with no files has no disks to buy and no traffic to pay for.

Even counted generously, and with the domain included, running NMTS costs me less than 50 dollars a month. That number leaves out the time I spend building it.

There is a donation feature, but it is optional, and right now the income is zero. It is still one of the most fun projects I have ever had.

So if donating is not for you, send me a bug or an idea instead. I will read it gladly.

needmoretruth, who are you?

Someone who really loves technology. Cryptocurrency most of all, and with it the things that come attached: cryptography, privacy and economics. I am not a cryptographer, not a genius programmer and not some hacker in the dark. I am a geek, and that is the whole description.

The name has a history that is both complicated and simple. When I picked a GitHub handle it was the first thing that came to mind, nobody else had it, so I kept it, and the matching domain happened to be free too.

There is no success story. I have loved computers since I was a kid, and philosophy and logic along with them. Mathematics I liked as an idea and disliked in practice, so I did not do much of it. I love ZK proofs, the kind that prove a thing without showing it, and I wanted them in NMTS, but there was no place where they belonged, so they are not in it.

My GitHub is full of things I built because they suddenly came to me. I daydream a lot, and when one idea sticks I write it in a notebook, and if it does not get in the way of NMTS, or I want it too badly, I build it. Usually my interest fades once the prototype and the skeleton are there, and I move on. NMTS is the first project I have carried this deep and this long without losing interest.

This may sound like a pose, but I am hopeless at marketing and money is not the goal. I do like money. With more of it I would build a dedicated server and spend a fortune on AI. I would not buy a supercar; to me the luxury life is being able to think and build what I thought of without having to work for it. Money is the means to that.

What are you really after? Why build this at all? Is it a honeypot?

I can see how it might look that way. There are several reasons, but the one at the centre is simple: this is fun for me. As I keep saying, I love technology. If I had to choose between technology and a relationship, I would choose technology, though I suppose one can have both. I do this because it interests me and I enjoy it, and I honestly doubt anyone would do this much for nothing otherwise.

Calling it volunteering is a stretch. But if someone stays up nights building a system, ships it, pays for a free trial, publishes the source and still earns nothing from it, then either they expect to earn far more later or the fun outweighs the loss. For me the second is the bigger part, with a little of the first.

A honeypot needs something to catch, and there is nothing here to catch. My server never holds a file it can read, you never gave it a name or an email, and the key never leaves your device. The parts you would have to take on faith are the published ones, and the recovery program gets your files back with no NMTS at all.

And who I am: I, needmoretruth, am Korean, born in Korea and living here. But going to the United States has been my wish since I was a child, and from high school on it was the first thing I meant to achieve in life. It still is: emigrating to the United States, to Texas, is my biggest goal.

If I get there, I will almost certainly, 99 % or so, change my legal name to Need More Truth, found a company called NeedMoreTruth, and move what I have been building in Korea under it. From where I stand the EB-3 skilled-worker route looks the most realistic; it is hard, and an O-1 would be wonderful if it were ever possible. Whatever the route and however many years it takes, I am going. If one day I have moved and read this answer again, I will be glad it is here.

Why are there no ads and no paid tiers?

Not because I am a nice person.

First, I hate carrying legal risk, and every ad network and every payment processor brings some.

Second, running NMTS costs very little. If it had cost a lot, I would not have made it paid; I would have found a way to make it cheap, or not built it at all.

Third, and this is the real one: until there are thousands of users, not charging is a better deal for me than charging. It keeps me further from legal trouble, and I am not going to hand my identity to an ad network to earn a few cents.

About Tusky…

I was lying in bed daydreaming, as I do, thinking about Sui, which I was into at the time. That led me to Walrus, and digging further I sketched a first NMTS prototype in my head, and digging further still I found out Tusky existed.

Tusky was an end-to-end encrypted storage on Walrus, and on Walrus it was the dinosaur. By its own numbers: 77.4 TB of raw data, 45.7 million files, 47 thousand uploading accounts in its last month alone. On 19 December 2025 it announced a shutdown, said the app, API and public aggregator would run through 19 January 2026, and later said on its X account that, with support from the Walrus side, the retrieval window was extended by 60 days to 19 March 2026. Billing stopped that day and December payments were refunded.

The official reason, in their words: it had demonstrated clear adoption but had not reached profitability, and fundraising had not progressed within the timeframe required. In plain words: it spent more than it earned.

What users were left with, per the retrieval guide, was downloading each storage space in turn or exporting blob IDs to move to another tool. The files themselves stayed on Walrus, but which pieces were yours and how to open them went through Tusky's app and API.

Meanwhile I had been chewing on three problems stacked together, legal, economic and technical, and what came out at the end was this: NMTS relays only metadata, and the files go between you and Walrus directly. And Tusky's situation is assumed from the start: keep a downloaded recovery list and the standalone recovery tool opens your files with no NMTS server at all.

NMTS uses none of Tusky's code, assets or trademarks. I still think of it as a fine predecessor that walked the road on Walrus first, at scale and bravely. Shutting down is a pity. I am a geek with no money, so I built NMTS to run on about 50 dollars a month, out of my own pocket, with no income.

If users suddenly multiply, will the server fall over or the bills explode?

Half right, half wrong. File bytes never pass through my server. Your browser encrypts the file, cuts it into Walrus pieces and sends them straight to a Walrus upload relay and the storage nodes. What the server gets per upload is two small requests: a registration record of a few hundred bytes and the index of your encrypted file list, a few KB. So however many users and files there are, the server's traffic and work stay tiny and do not depend on file size.

In numbers, as of 6 September 2026 at midnight UTC: on the same machine as this server, running the same server image in the test stack, it took up to about 930 upload registrations per second with no failures. So 500 people each uploading a 3 MB photo every second is comfortably within that.

It is not perfect. At several times that, answers slow down and the per-account and per-address request limits answer first with "try again shortly" (429). The server queues rather than dies. Peaks are not sustained either. Files uploaded with credits carry one chain transaction each, so they are slower and have a daily cap.

The bill does not multiply with requests. Cloudflare stands in front and the server is my own local machine rather than a VPS, so ten times the requests is not ten times the invoice. I think it likely stays under 50 dollars a month.

One "upload" was replayed exactly as the server receives it: one file registration request (POST /v1/items) and one write of the encrypted file list's index, 4 KB (PUT /v1/manifest). File bytes never reach the server, so they are not part of it.

First run: 700 accounts, 24 concurrent requests, 10 seconds. 8,663 registrations, 864 per second, 0 failures, 16 ms and 12 ms per request on average. Second run: 1,300 accounts, 48 concurrent requests, 10 seconds. 9,308 registrations, 929 per second, 0 failures, 29 ms and 23 ms. Doubling the concurrency left the rate almost unchanged and doubled the latency, so that is about the ceiling of this setup.

During the runs the server process used about 1.6 CPU cores and the database about 3.7. What sets the ceiling is not the CPU but the 10 connections the server opens to the database. The live server and the test stack share this machine, so the sentence above uses 500, a little over half of what was measured.

Why a local server instead of a VPS (AWS, GCP and the like)? Is that not risky?

The cheap answer first: I have no money. A stable, predictable bill I could live with, but a setup where a traffic spike melts my money is more than I can carry, and a VPS is too expensive for me. I have always been at home on Linux, so NMTS runs on a mini PC with Ubuntu Server.

The risks are power and the disk. In a power cut it stops until power returns, then the containers bring themselves back (restart policy: always). If memory runs short, a watchdog kills the largest process before the whole machine freezes. And the power cord is secured so it does not get pulled.

Backups: the database goes to Cloudflare R2 every day at 03:30 UTC (locked for the first 7 days so it cannot be deleted), plus an hourly copy whenever something changed. I have actually restored from it, on the day the backup was made. And if the server is down, your files are still on Walrus; with a downloaded recovery list the recovery tool opens them without the server. What the server could lose is the encrypted lists, and those are what the backups hold.

If money or real income ever arrives, moving to a VPS is something I seriously consider.

Why Cloudflare? You said you cared about privacy.

I bought the domain at Cloudflare, and a Cloudflare Tunnel hides my IP (there is no port forwarding). The core of it: I have no power to stop a DDoS. Putting Cloudflare in front filters out countless attacks by itself, it is fast, it caches well, and it is free. The human check (Turnstile), the lock in front of the admin screen (Access), the backup store (R2) and the watchdog that checks the server from outside all run there too.

It is one more third party to trust, yes. TLS ends at Cloudflare, so Cloudflare sees what my server sees: ciphertext, counts, sizes and times, and your address. File bytes go from your browser straight to Walrus and never pass Cloudflare. Keys pass nowhere.

Why I think that is acceptable, two reasons. What passes is already ciphertext, so nothing opens because Cloudflare saw it; and Cloudflare publishes government-request statistics in a transparency report and has a published policy of notifying customers unless the law forbids it. Without Cloudflare I would have to open a port, expose my IP, and that IP would be eaten alive by DDoS.

You are one person. If you stop developing, or switch the server off, what happens to my files?

They stay where they are. Your files are not on my server. They are on Walrus, a public storage network, encrypted, for the period that was paid for. My server going away does not remove a byte of them.

Getting them back without me is what nmts-recovery is for: one open-source program for Linux, macOS and Windows. Give it your NMTS key and your recovery list and it fetches the pieces from public Walrus nodes, decrypts them and writes your files out. It contacts no NMTS server at any point, because it needs nothing a server has: every key comes from your NMTS key, and where each piece sits is in the recovery list. It uses the same published encryption engine as the browser, so a file encrypted here opens there.

The recovery list is the one thing to keep. You can download it from the account screen, and you can switch on a copy on the storage network itself: for the uploads you make after switching it on, your NMTS key alone is then enough, because the program derives your wallet address from it and reads the list from a public node, with nothing of NMTS left standing. The copy is a switch rather than the default because it rides in an upload you pay for.

One limit, plainly: files paid for with credits sit on storage the donation pool bought, and you cannot extend that period yourself. Files paid from your own wallet belong to that wallet on the chain, and their period can be extended by whoever holds that wallet.

If the recovery program works without any server, what is NMTS for?

For every day. Walrus stores bytes; it does not encrypt them, name them, keep folders, send one file to one person, take a payment in credits, or open on a phone. NMTS is that layer: the encryption, the file list, sharing, the wallet and the credits, the screens and the terminal tool. The recovery program is the exit, not the product.

Why the exit exists at all: as you know by now, NMTS is not run by a genius developer or a rich businessman, and people rightly dislike being tied to something like that. So the parts you would otherwise have to take on faith are open source, and the recovery program removes the single point of failure, which is me. That seemed to me the only way a project like this survives while opening up ground nobody had taken.

What is your vision?

In one sentence: NMTS becomes the single giant that takes the whole drive market. I am one developer, so that road does not come from my hands but from an SDK that lets people put NMTS into their products easily, and from a base people can build a business on with NMTS as the backend.

For storage I added Filecoin to Walrus as NMTS Heavy in September 2026, and I plan to add Arweave. The fee is 0 % and NMTS stays non-profit. The long version is on the vision page.

How can I help?

With feedback, more than with money. Found a bug? Report it, however small, and even if you are not sure it is one. Where the buttons sit and how the screen looks count too. Missing a feature, or dislike one? Tell me. And if you just want to say something, send an inquiry about anything, or about nothing in particular.

The code that matters is public: the CLI, the SDK, the encryption engine and the recovery tool are on GitHub, and a bug report there works just as well. There is a Discord server, and I am on X.

If you do want to give something, you can do it from the Wallet screen in the app, and the hall of fame lists everyone who did. Nothing is owed for it and nothing comes back for it; the terms say so in full.