PARS Preview

Working MVP · Event Preview v0.3

Built for Trust

Enterprise AI Governance & Transformation Platform

Gain visibility, reduce AI risks, understand costs, and empower your organization to adopt AI responsibly.

Explore Ecosystem View Roadmap
  • Visibility first
  • Vendor-neutral
  • Privacy-first
  • People-first

The PARS ecosystem — six connected modules

Select a module to see its scope and status.

Milestone

Completed (manual test)

First end-to-end enterprise integration verified

A user signed in with Microsoft Entra ID sent a test event from the PARS extension through the Gateway API to MongoDB, verified in a manual integration test.

  • Extension MVP
  • Entra ID sign-in
  • Gateway API
  • MongoDB

Automatic forwarding of real prompt events to the Gateway is planned.

Architecture & verification details

Product Roadmap

From Visibility to AI Transformation

A phased journey toward a trusted enterprise AI ecosystem.

  1. Phase 01

    Visibility & Governance

    In Development

    Organization-wide visibility into AI usage, identification of potential risks, and governance insights.

    Current scope

    • Browser Extension
    • Risk Dashboard
    • Gateway / DLP
    • Organization Visibility
    • Microsoft Entra ID SSO integration

  2. Phase 02

    AI Policy Builder

    Planned

    Define organizational AI policies, guardrails and usage rules.

    Planned capabilities

    • Policy templates
    • Organization-wide guardrails
    • Approval workflows

  3. Phase 03

    AI Cost Center

    Planned

    Understand AI usage costs, consumption patterns and optimization opportunities.

    Planned capabilities

    • Usage tracking
    • Cost analytics
    • Chargeback & reporting

  4. Phase 04

    Prompt Library

    Planned

    Reusable, approved, high-quality prompts and organizational knowledge sharing.

    Planned capabilities

    • Curated, approved prompts
    • Organizational knowledge sharing
    • Quality & risk review

  5. Phase 05

    Agent Management

    Planned

    Govern enterprise AI agents through lifecycle management, access controls and visibility.

    Planned capabilities

    • Agent inventory & visibility
    • Lifecycle management
    • Access controls

  6. Phase 06

    AI Learning & Maturity

    Future Phase

    Improve AI literacy and responsible usage, and measure organizational AI maturity.

    Planned capabilities

    • AI literacy learning paths
    • Responsible-use guidance
    • Maturity assessment

Phases show the planned sequence. Module status describes each module as a whole; individual components are tracked separately under Development. No release dates are committed at this stage.

Phase 01 delivery steps

Visibility & Governance: what is done and what comes next.

Completed

  1. Extension MVP / Firestore event capture
  2. Gateway API foundation
  3. Entra ID PKCE authentication
  4. Gateway JWT / Edge.Write authorization
  5. Gateway → MongoDB event persistence
  6. End-to-end integration verificationCompleted (manual test)

Next stepsPlanned

  1. Firebase Only / Hybrid / Gateway Only runtime modes
  2. Automatic Extension → Gateway event forwarding
  3. Centralized DLP and policy enforcement
  4. Enterprise dashboard integration
  5. Organization onboarding and a controlled pilot

Currently Building

Turning Vision Into Reality

The first module, Visibility & Governance, is being built through four engineering workstreams. The architecture below shows what works today and what is planned next.

Architecture

Two flows, verified separately

The working MVP flow and the enterprise integration path run separately today. They are not yet connected into one automatic flow.

Working MVP flow

Completed
Chrome Extension Captures ChatGPT prompt events
Firebase Firestore Provider, device, session and risk data

Enterprise integration path

Completed (manual test)

Identity & authorization layer

Microsoft Entra ID Authorization Code + PKCE sign-in · access token with Edge.Write
Chrome Extension Sends a manual test eventSigns in with Entra ID (PKCE)
PARS Gateway API .NET · receives and stores eventsValidates the Entra JWT and Edge.WritePlannedCentral DLP & policy enforcement
MongoDB edgeEvents collection

Planned next

  • Automatic forwarding of prompt events from the extension to the Gateway
  • Runtime modes: Firebase Only / Hybrid / Gateway Only
  • Central DLP and policy enforcement in the Gateway
Status legend Completed Completed (manual test) Planned
Verification results (manual integration test)
202 Accepted
Authorized request with a real Entra ID access token that includes Edge.Write.
401 Unauthorized
Request without a token was rejected, and no event was stored.
403 Forbidden
Valid test token without Edge.Write, checked in an integration test with a mock token issuer.
MongoDB
The test event was stored in the edgeEvents collection and checked directly in the database.
Firebase
The existing Firebase capture flow was confirmed to be unaffected.

These results come from manual verification. They are not a production security assessment.

Engineering workstreams

  • Working MVP

    Browser Extension

    Captures prompt events in ChatGPT and records them in Firebase Firestore.

    Current focusAutomatic forwarding of prompt events to the Gateway is planned.

    Technical details

    Built so far

    • Prompt event capture in ChatGPT
    • Event records in Firebase Firestore with provider, device, session and risk information
    • Microsoft Entra ID sign-in (Authorization Code + PKCE)
    • Manual test event sent to the Gateway API
    • Access token kept only in service-worker memory, not persisted

    Planned next

    • Automatic forwarding of prompt events to the Gateway
    • Runtime modes: Firebase Only / Hybrid / Gateway Only

    The existing Firebase flow was confirmed to be unaffected by the new integration.

  • Working Prototype

    Risk Dashboard

    Visualizes risk levels, AI usage patterns and organizational insights.

    Current focusThe current runnable scope of the prototype is being confirmed.

    Technical details

    Built so far

    • Existing dashboard implementation for risk levels and AI usage insights

    In progress

    • Confirming the current runnable scope

    Planned next

    • Enterprise dashboard integration
  • Working MVP

    Gateway / DLP

    Server-side entry point that checks identity and permissions before storing events.

    Current focusThe Gateway API is working. Central DLP and policy enforcement are planned.

    Technical details

    Built so far

    • .NET Gateway API running
    • Entra ID JWT validation and Edge.Write authorization
    • Event persistence to MongoDB (edgeEvents)

    Planned next

    • Central DLP and policy enforcement
    • Automatic event intake from the extension
  • Verified

    Microsoft Entra ID SSO

    Enterprise sign-in for the extension and token-based authorization for the Gateway API.

    Current focusVerified for PKCE sign-in and Gateway authorization. Organization onboarding is planned.

    Technical details

    Built so far

    • Authorization Code + PKCE sign-in with a real user from the extension
    • Access token for the Gateway API with the Edge.Write scope
    • Gateway authorization checked: 202 with Edge.Write, 401 without a token, 403 without Edge.Write (mock-issuer test)

    Planned next

    • Organization / tenant onboarding

    Verified scope: authentication and Gateway authorization only. Organization onboarding is not complete.

Component statuses describe engineering progress, not general availability or production readiness. Visibility & Governance remains In Development until automatic event forwarding, central DLP and dashboard integration are complete.

About the Project

Responsible AI adoption, built step by step

PARS is being developed to help organizations adopt artificial intelligence with greater visibility, confidence, and responsibility.

The platform is being built incrementally. Governance and visibility come first: an organization needs to understand how AI is actually used before it can set meaningful policies, manage costs, or measure maturity. Each following module builds on that foundation.

PARS aims to build a culture of safe and informed AI use, not to monitor individual employees. It helps organizations understand AI usage patterns and risks, and guides teams toward safer practices within clear, organization-level guardrails.

Built by

Gökhan Kahraman

gkahraman.com (opens in a new tab)
PARS Module