// WRIT // SOVEREIGN AI BACKEND · // NIST 800-53 REV 5 // OSCAL COMPONENT DEFINITIONS · // HYBRID X25519+ML-KEM-768 TLS · // 100% APACHE / MIT / BSD / MPL · // CNSA 2.0 ALIGNED · // ONE OPENAPI CONTRACT · // IL4 / IL5 TARGET · // AIR-GAP READY · // WRIT // SOVEREIGN AI BACKEND · // NIST 800-53 REV 5 // OSCAL COMPONENT DEFINITIONS · // HYBRID X25519+ML-KEM-768 TLS · // 100% APACHE / MIT / BSD / MPL · // CNSA 2.0 ALIGNED · // ONE OPENAPI CONTRACT · // IL4 / IL5 TARGET · // AIR-GAP READY · // WRIT // SOVEREIGN AI BACKEND · // NIST 800-53 REV 5 // OSCAL COMPONENT DEFINITIONS · // HYBRID X25519+ML-KEM-768 TLS · // 100% APACHE / MIT / BSD / MPL · // CNSA 2.0 ALIGNED · // ONE OPENAPI CONTRACT · // IL4 / IL5 TARGET · // AIR-GAP READY · // WRIT // SOVEREIGN AI BACKEND · // NIST 800-53 REV 5 // OSCAL COMPONENT DEFINITIONS · // HYBRID X25519+ML-KEM-768 TLS · // 100% APACHE / MIT / BSD / MPL · // CNSA 2.0 ALIGNED · // ONE OPENAPI CONTRACT · // IL4 / IL5 TARGET · // AIR-GAP READY · // WRIT // SOVEREIGN AI BACKEND · // NIST 800-53 REV 5 // OSCAL COMPONENT DEFINITIONS · // HYBRID X25519+ML-KEM-768 TLS · // 100% APACHE / MIT / BSD / MPL · // CNSA 2.0 ALIGNED · // ONE OPENAPI CONTRACT · // IL4 / IL5 TARGET · // AIR-GAP READY · // WRIT // SOVEREIGN AI BACKEND · // NIST 800-53 REV 5 // OSCAL COMPONENT DEFINITIONS · // HYBRID X25519+ML-KEM-768 TLS · // 100% APACHE / MIT / BSD / MPL · // CNSA 2.0 ALIGNED · // ONE OPENAPI CONTRACT · // IL4 / IL5 TARGET · // AIR-GAP READY ·
§ FOR DEVELOPERS

Vibe-code
mission-critical
apps.

Writ is the only sovereign AI backend built to be vibe-coded against. Your developers describe what the mission needs in plain language to the AI coding assistants they already use — and a working app runs against real data, inside your fence, in hours. One toolkit. One audit trail. One accreditation boundary that doesn't break when the pace picks up.

§ ONE WAY IN

Install the toolkit. Ask Writ something.

Writ speaks the same standard web language that popular AI services already speak. Developers use a familiar toolkit, call a simple command, and get an answer.

The same toolkit works for every kind of AI — chat, prediction, search, image understanding, agents. Learn one, know them all.

// install the toolkit
writ
mTLS · hybrid-X25519+ML-KEM-768
 $ pip install writ
// or, in node
writ
mTLS · hybrid-X25519+ML-KEM-768
 $ npm install @writ/sdk
// ask a question
writ
mTLS · hybrid-X25519+ML-KEM-768
 $ writ ask "summarize this document"
§ FITS YOUR EDITOR

Plays well with Cursor, Claude Desktop, and the rest.

Your developers' editors can plug into Writ with a single command. The assistant sees only what your organization allows it to see. Every call is authenticated. Every call is logged.

Give an editor the shape of the Writ API and a prompt like "scaffold a claims-review app against our internal data" — and the editor will generate a working starting point in minutes.

◆One-line install for popular AI editors.

◆Uses your identity system — single sign-on included.

◆Scoped permissions per developer, per tool.

// connect Claude Desktop to your Writ
writ
mTLS · hybrid-X25519+ML-KEM-768
 $ writ connect claude-desktop
// or Cursor
writ
mTLS · hybrid-X25519+ML-KEM-768
 $ writ connect cursor
§ REFERENCE APPS

Working starting points.

We publish working reference applications alongside every major release, so a developer new to Writ can read real code against a real backend and see how the pieces fit.

01 ~3 hrs

Document triage

Classify inbound documents, pull key facts, route to the right team, draft a summary with citations.

02 ~4 hrs

Anomaly watch

Forecast trend lines over operational telemetry, flag unusual readings, draft an incident one-pager.

03 ~6 hrs

Analyst assistant

A desktop assistant over your document corpus, with citations, image understanding, and per-paragraph risk scoring.

§ FOR TECHNICAL READERS

Toolkit reference, API shape, error model — all in one paper.

Read the Architecture Paper