Preview · Plena AI — coming soon

Proof Readiness Assistant.

PlenaProof is not adding AI to become another chatbot. PlenaProof is adding native AI to help people and institutions convert messy real-world claims into structured, reviewable, receipt-backed proof workflows. The AI assists preparation, classification, evidence-gap analysis, receipt drafting, and human accountability. It does not replace verification, institutional authority, or human review.

Preview of Plena AI — coming soon

This is an early preview of Plena AI, PLENA's own forthcoming accountability AI, which will run on open-source models PLENA hosts on its own infrastructure — for independence from commercial AI providers, rather than reliance on their APIs. For now it runs entirely on your device in demo mode — nothing is sent to any third-party AI service. It demonstrates the proof-readiness workflow; it is not a finished product, and it is not a general chatbot.

Describe what you are trying to prove

AI can generate answers. PlenaProof helps structure what can be reviewed, recorded, and trusted.

Try an example:

This preview runs entirely on your device — nothing is sent to any third-party AI service. Plena AI — PLENA's own accountability AI, built on open-source models it runs on its own infrastructure — is coming.

Truth boundary. PLENA AI provides a verification-readiness and evidence-organization summary. It does not perform official verification, authenticate identity, confirm document validity, or replace legal, institutional, or human review. Any receipt generated here is draft or self-declared unless issued by an authorized institution through a real review process.

What this preview is

An honest demo of the workflow today, and a clear statement of what Plena AI will be.

Today

On-device demo

The preview runs entirely in your browser. It maps your inputs to a structured proof-readiness summary and a draft receipt — instantly, offline, with nothing sent to any server or third-party AI.

Coming

Plena AI

PLENA's own accountability AI is a future endeavour, designed to run on open-source, open-weight models (open-weight families such as Llama, Mistral, Qwen, or DeepSeek) on infrastructure PLENA controls — independent of commercial AI providers, so access, pricing, and terms stay in PLENA's hands. When it arrives it will power richer analysis under the same truth boundary — assist and organize, never verify on its own.

Always

The truth boundary

In every form, now and later, the assistant prepares and organizes. It does not perform official verification, authenticate identity, or replace human or institutional review.

Offline drafts. Offline drafts help you prepare information when connectivity is limited. They are not official PLENA receipts and are not submitted for review until synchronized or processed online. Only simple draft text is stored locally — never identity, medical, immigration, or legal documents.

Built for institutions

Pilot it with your own controls — own-key data sovereignty, audit trails, and a clear truth boundary.

Your key, your backend

Route through your own secure backend and model account (Mode B). PlenaProof never stores your model keys, and an optional API-key gate restricts who can call the endpoint.

Audit & observability

Every response carries a request ID. Optional metadata-only audit logging, per-IP rate limiting, request timeouts, rate-limit headers, and a health endpoint for your ops team.

Honest by design

The assistant prepares and organizes; it never claims official verification. That truth boundary is enforced both in the interface and on the server.

How your data is handled

Plain, enterprise-grade data handling. No surprises.

What is sent

In Mock/demo mode, nothing leaves your device. In live modes, only the text you typed (claim, who must trust it, evidence notes) is sent to the configured AI provider to produce the summary. Documents and files are never uploaded.

What is stored

Only a local draft of your typed text in your own browser, which you can clear anytime. The server keeps no copy of your claim or evidence; optional audit logs record metadata only (timestamp, outcome, latency) — never your text.

Your controls

Use Mock mode to stay fully on-device. Use Clear to wipe the local draft. Do not paste government IDs, card numbers, or medical records — you will be warned if the assistant detects them before any live send.

Processing basis & retention. Live analysis is processing you initiate for your own preparation. Inputs are not used to train PLENA models and are not retained server-side beyond the request. Where enabled, audit logs are metadata-only and subject to a retention window. This tool is not a controller of record for any institution; institutions deploying it should complete their own privacy review and route requests through their own backend (Mode B).