Storyline / Showcase
Write the work
you can't paste.
A private writing studio for people whose words are confidential by law: clinicians, lawyers, journalists, anyone who cannot put sensitive work into a cloud chatbot. The AI runs on your device, nothing leaves unless you approve it, and every crossing is on the record.
Try it: the door out
A blocked search is the product doing its job.
Every web search in Storyline passes one door. It is screened against the project's protected names first, you see the exact text that would leave, and nothing goes until you choose. Try searching from a confidential draft.
Select a passage and choose Search the web.
What left this device
- Nothing yet.
What stayed on the device
- Nothing yet.
A simulation in your browser: nothing here touches the network. In Storyline the screening, the generalising (by the on-device model) and the ledger all run on your device, every crossing is sealed into the project's Data flow ledger and hashed into the audit chain, and an on-device guard also warns when a run of clean searches starts circling a protected topic.
Where the AI runs
Privileged work stays on the device. In code, not a setting.
Every model call goes through one resolver. Each project carries an access policy, and a privileged project is forced to the strictest one before any request is built.
Any provider
Apple Intelligence, an MLX model on the device, your own Ollama, a model on another of your devices, or a cloud provider with your own key. The default for ordinary projects.
Your devices only
This device, or a trusted device of yours on the local network. Third-party cloud is refused.
This device only
Enforced for every privileged project. Cloud, custom endpoints and even your own nearby devices are refused. With no local model, the request fails with a clear message; there is no cloud fallback.
Provable, not promised
A log you can check, even against me.
Your work sits in a vault: the key comes from your passphrase through Argon2id, and content is sealed with AES-GCM, bound to its context. Every time confidential content is opened, exported or sent to a model, Storyline writes an entry to a hash chain signed by a key that never leaves the Secure Enclave. Edit, delete or truncate an entry and the chain breaks at that point.
Anything derived from privileged content inherits its protection or is never stored: snapshots and AI outputs are sealed, research threads live only in memory, and search history is sealed to the vault.
- Vault unlockedhash ← previous hash + entrysigned in the Secure Enclave
- AI run, on devicehash ← previous hash + entrysigned in the Secure Enclave
- Search approvedledger entry hash includedsigned in the Secure Enclave
- PDF exportedfile SHA-256 includedhead pinned outside the file
Don't trust it. Test it.
The airplane test.
Before every external build, writing and AI have to work with no network at all. You can run the same test yourself.
- Airplane Mode onUnlock the vault with Face ID or Touch ID, offline.
- WriteOpen a privileged project and type. Nothing hangs.
- Ask the co-authorA branch arrives from the local model.
- Check the logThe unlock, the AI run and the merge are there, with hashes.
- Back onlineSync resumes, with no duplicates.
All nine steps, and how every claim is mapped to the code that makes it true →
Behind the build
From the journal
My research agent cited example.com
Getting a small on-device model to actually do research, and a ledger that shows exactly what left the device.
I tried to search the web from my own app. It said no.
Storyline 1.2.1: protected names are held back, the query is rewritten on device, and the research still comes back cited.
Your own devices as a private AI cluster
The most capable AI computer you own is probably asleep in the next room. Storyline's next bet is to use it.
Show me: what happens to a sentence you write in Storyline
No server, an Argon2id vault and a tamper-evident audit log you can verify yourself. The machinery behind provable privacy.
Promised versus provable
Private AI is becoming a marketing phrase. For confidential work, privacy has to be something you can check, not something you are told.
Siri can call Google now. Where does confidential work go?
If Apple sends its hardest questions to a partner cloud, "it runs on my device" is no longer a given. For some professionals, that is a dealbreaker.
Free to download.
On iPhone, iPad and Mac. Your words stay yours.