LIBRA OS // v0.1.15
Runtime online Libra OS / Packs / Practice

A supervised AI workforce, running on your judgment.

Libra OS compiles how you actually practice into an expert PROFILE, then runs it across a team of supervised twins — intake, drafts, conflict checks, follow-ups. Each twin acts below a confidence + scope line and halts above it. Nothing ships without your sign-off.

PROFILE :: process table  4 twins · supervised
01
intake.twin
prospect calls · web-chat · conflict scan
running
02
drafting.twin
engagement letters · forms · routine docs
running
03
status.twin
matter inquiries · back-end lookups
running
04
matter.twin
deadlines · doc-chasing · status drafts
running
01

The front office is the bottleneck.

Intake eats the hours you'd spend on work that needs judgment. Conflict checks get sloppy at volume. Drafts are templates you fill by hand. Routine status questions chew through afternoons.

Hiring solves it badly: high overhead, long training, and most of what an assistant does is repetitive enough that a well-supervised AI should handle it.

A general copilot helps one professional draft faster. It doesn't multiply a practice — it doesn't run the front office, hold the line on scope, or keep the file moving while you focus on the work that needs you.

02

Four supervised twins. One PROFILE.

Feed it 30–50 real intake conversations, a few engagement letters, and your principles for sensitive cases. Libra OS distills that into your expert PROFILE — then runs it across four twins, each with its own autonomy threshold.

TWIN 01 · FRONT DESKauto

Intake Twin

Handles prospect calls and web-chats. Collects facts, runs your conflict scan against the client list, books consults. Hard-stops the instant a prospect asks something that needs your judgment.

autonomy
72%
TWIN 02 · DOCUMENTSauto

Drafting Twin

Pre-fills engagement letters, intake forms, and routine documents from facts collected. Flags any field it can't fill and surfaces missing inputs. You review, customize, sign.

autonomy
64%
TWIN 03 · STATUSguarded

Status Twin

Answers routine “where's my matter?” inquiries from your back-end. Escalates the moment the answer needs nuance — bad news, requests for more info, anything that changes strategy.

autonomy
58%
TWIN 04 · MATTERauto

Matter-Management Twin

Watches deadlines. Nudges clients for missing documents. Drafts routine updates for your review. Keeps the file moving while you focus on the work that needs you.

autonomy
80%
03 — THE HARD-STOP, LIVE

The twin runs the routine. Then halts and hands you the judgment.

Below the line, autonomous. The instant a question crosses into your professional judgment, the process freezes and escalates — every word logged to your PROFILE.

intake.twin · stdoutprospect · web-chat
twin is responding…
supervisor desk0 awaiting
No escalations. The twin is handling routine intake autonomously.
audit trail
04

One substrate. Vertical packs on top.

The regulated work of a professional rarely changes shape across fields. What changes per vertical: intake taxonomy, templates, scope boundaries, compliance overlay. Packs ship alongside the platform.

LAW Immigration · family · PI · IP · estate · criminal FIN CPAs · tax prep · bookkeeping ADV Fiduciary advisors · planners CARE Therapists · counselors · coaches CONS Management · IT · HR consulting + Architecture · engineering · your field

Each pack ships a regulator-defensible default policy — UPL for law, AICPA scope for accounting, HIPAA-aware gating for care, SEC/FINRA suitability for advisory.

05

How it stays your license, not the AI's.

Three guardrails, compiled in — not optional. Defensible to your bar, board, or insurer.

GUARDRAIL 01

Hard-stop on out-of-scope

Questions demanding your judgment — “do I qualify?”, “what should I do?”, “is this the right treatment?” — never get answered by the AI. They escalate. The threshold can't drop below your field's regulatory floor.

GUARDRAIL 02

Threshold supervision per task

Each twin acts autonomously below a confidence + scope line and escalates above it. The default is regulator-defensible for your vertical; you tune within bounds, never below the floor.

GUARDRAIL 03

Audit-logged & attributable

Every word a twin sends is logged and attributable to its task and your PROFILE. If your bar, board, or insurer asks “what did the AI say?” — you have the trail.

06

Discipline first. Surface second.

The Practice pack is built by Nebula Nova Inc. on the Libra OS engine — supervised professional-services infrastructure. The discipline mechanism — UPL-style scope guardrail, conflict scan, audit logging, supervision spectrum per task — was first validated against a PGWP immigration pack, where the regulatory bar is among the strictest.

EqualDocs is our first design partner; we're piloting the supervised-twin substrate with them on SME contract review and intake. Different vertical, same discipline architecture.

We spent the last year building the substrate — scope guardrail, PROFILE distillation pipeline, supervised-agent runtime — before opening the surface. The substrate is ready for the second professional. The product is ready for the tenth.

07 — REQUEST ACCESS

The next ten get concierge setup.

Tell us about your practice. We'll distill your conversations into a PROFILE, tune the threshold policy for your vertical, and walk the audit trail end-to-end.

libra-os · onboarding/request

No marketing emails. Onboarding contact only.