macOS Technical Preview

Trulana

Give AI context —
without giving it your data.

Trulana is a local-first private context server. Apps query local context over REST or MCP, and only redacted answers leave the device.

REST + MCP verified · 117 tests passing · Runs locally
Encrypted local vault Localhost-only REST MCP support Audit log + redaction
What is Trulana?

See the full flow in under a minute.

Authenticate an agent, run a live query, watch redactions apply, and see the audit log update in real time.

Dashboard
Local vault running strict
Connected Agents
No agents connected
Audit Log
Waiting for requests...
trulana localhost:8432

AI that doesn't know you is useless.
AI that knows everything is dangerous.

Right now you have two choices. Give your apps zero personal context and get generic answers. Or hand over your emails, calendar, health data, and notes — and hope nobody leaks it, trains on it, or sells it.

Option A

No context

Every session starts from scratch. The AI doesn't know your name, your schedule, or what you asked yesterday.

Option B

Full access

Send everything to the cloud. Your data trains models, populates dashboards, and lives on servers you don't control.

Apps get answers, not your data.

Trulana stores your personal context in an encrypted local vault. When an app needs to know something about you, it queries Trulana — and gets back a filtered, redacted response. Your name, email, location, and schedule never leave the device in raw form.

The app gets useful context. You keep control of the source material.

Five steps. All local.

01

App requests access

An app sends an auth request with its name, ID, and stated intent.

02

Trulana issues a scoped token

Short-lived, in-memory, 15-minute TTL. Doesn't survive a restart.

03

App queries your context

"What's their morning routine?" — sent to localhost, never to the internet.

04

The redaction engine runs

Three stages: regex PII stripping, entity recognition, privacy-level generalization. Emails, phones, SSNs, locations — all caught.

05

Filtered result returned

The app gets useful context with all identifying details stripped. The raw data never moved.

Built for people who build with AI.

Developers building AI apps

Need user context without storage liability? Query Trulana via REST or MCP. Get personalization without the GDPR headache.

Consultants & professionals

Use AI with client data, NDAs, and sensitive materials — without sending any of it to a cloud provider.

Privacy-aware power users

Get the "AI that knows you" experience without the tradeoff. Your context, your device, your rules.

The current model doesn't scale.

Every AI app that stores personal context becomes a data liability. One breach, one policy change, one acquisition — and your users' data is somewhere it shouldn't be.

Trulana eliminates this by never collecting the data in the first place. The context lives on the user's machine. Your app queries it. The redaction engine enforces the boundary. You get personalization without the storage, the GDPR headache, or the trust problem.

Working macOS MVP. Build it and run it today.

$ git clone https://github.com/AdamsLocal/trulana && cd trulana && flutter run -d macos COPY
macOS app with encrypted local database
REST API on localhost:8432
MCP integration (works in Cursor today)
Token-based auth with 15-minute TTL
3-stage auto-redaction pipeline
Encrypted audit log for every query
117 tests passing, 38 REST + 18 MCP checks
Scope enforcement (in progress)
On-device LLM for NER (next phase)

Try the architecture. Then decide if you want to build on it.

I'm looking for a few macOS design partners building AI workflows that need local-first context and redacted outputs.