Back up and export the workspace
Workspace owner · Safeguard the work
Why this matters
Your workspace is a database on a relay, and if that relay is lost, so is everything in it. Portability exists in principle — the data is signed events in an open format and your identity is yours — but principle is not a restore.
Steps at a glance
1. Establish what backup actually exists for your relay
Ask the question directly of whoever runs the relay: what is backed up, how often, where is it stored, and when was a restore last tested.
If you are on a hosted relay, this is a supplier question and the answer belongs in your project risk record. If you self-host, it is your answer to give.
Warning
In current versions there is still no user-facing workspace or message export — nothing in Settings. Re-verified against source at v0.5.18 (2026-08-22) and again at v0.5.20 (2026-08-26): unchanged. Re-check with each release. The local-archive settings now default to enabled, but that is a local cache, not an export you can take somewhere.
2. Separate the two things worth preserving
Your identity is your keypair. The app can produce an encrypted backup of it, with a passphrase and a test-restore flow — so this half is solved and takes five minutes. See Back up and protect your private key.
The workspace content is messages, threads, canvases, media and the audit log, held by the relay.
These fail independently. Losing the relay with your key intact means you still are who you are, in an empty world. Losing your key with the relay intact means the world is there and you cannot enter it.
3. Get a copy out
Two kinds of copy, and the difference matters.
A physical backup is the relay's underlying storage — a database and blob backup. It restores the whole thing exactly, and it is only usable by someone who can operate the relay stack.
A logical export is the events themselves in open signed-JSON form. It is far more portable — readable by other Nostr tooling, independent of any particular relay implementation — and it is the thing that makes "you can leave" more than a slogan.
The planned managed-hosting design treats a continuous logical export as the exit hatch precisely because it survives the vendor. Whether the shipped product offers anything equivalent today is the open question in step 1.
4. Verify the copy
An unverified backup is treated as no backup. That is not a slogan either — it is the practice the design doctrine explicitly adopts, with automated restore-verification on a schedule.
At minimum: confirm the copy is complete, confirm it can be opened, and confirm someone other than its creator could use it.
5. Know your recovery position
Write down, for each active project, what happens if the relay is lost tomorrow: what is recoverable, from where, by whom, and how long it takes. If the plain answer is "nothing", that belongs in the project's risk record and, arguably, in a conversation with the client.
Also note the softer failure: asynchronous replication can lose the last seconds of writes. Not usually material for project work — but it means "backed up" and "identical" are not the same claim.
Verified against Buzz v0.5.20 · Updated 2 Sep 2026