Problem statement

GST registration involves collecting business details and documents, checking them, then entering the same information on the government portal. Separate collection and filing steps create repeat entry and follow-up work. Bottlenecks at three places being getting complete data from client, availability and execution promptly by staff and timely approval by CA engaged in preoccupancy together add to the delay in service reciepe.

Solution overview

A web application brings client intake, document upload, application review and assisted GST portal entry into one workflow. It also prepare the NOC and authorized signatory drafts for signature based on data provided. The practitioner remains responsible for checking the information and completing required verification.

Technology used

HTML, JavaScript; Node.js and Express; Supabase; Socket.io; Playwright with Chromium; Docker. Built with AI-assisted development using Google Antigravity.

Role of AI

The product uses rule-based browser automation (RPA). No runtime AI model, LLM, OCR or machine-learning service was found in its main execution paths. AI was used in development: the author built the code with Google Antigravity, Google's agentic development platform.

Expected impact

Less repeated typing, one place to review client information and clearer visibility of automation progress. Multi state Registration can be done by editing one application, time saved on portal errors and glitches, no losing unsaved data as any detail entered in system is autosaved.

Implementation workflow

1. Client signs in and enters registration details through a multi-step form, with supporting documents.
2. Application data is sent to the API and stored in Supabase for review.
3. Practitioner reviews the application and uses the accept/reject and automation controls.
4. Backend starts the Playwright runner. Socket.io returns progress and browser screenshots to the review screen.
5. Runner contains GST Part A, CAPTCHA/OTP handling, TRN capture and Part B data/document filling paths.
6. CAPTCHA and OTP answers come from the human operator. The code then requests review and submission using DSC or EVC; this is not proof that final filing succeeded.

Implementation scope

The code connects client intake to professional review and a browser-driven portal workflow. It is not an official GST API integration and does not make tax or eligibility decisions.

Data and documents

The backend uses Supabase application records and includes local document-file handling. Current code also writes to a second Supabase instance. Document storage and retention need a controlled design before client rollout.

Human in the loop

Professional review, CAPTCHA/OTP entry and final portal verification remain human steps. The final submission message is implemented in the runner; access to a usable final submission session and real filing completion still need separate verification.

AI boundary

Playwright is browser automation, not an AI model. AI enters through development.

Impact and evidence

Impact measurement

Measure manual preparation time and assisted preparation time for comparable cases, corrections after review, and completed filing outcomes. Until those measurements exist, benefits remain expected rather than quantified.

What is not claimed

No guaranteed typo-free filing, autonomous submission, live-portal reliability, GSTIN approval, successful production deployment or measured saving is asserted.

Before client rollout

Reconcile repository and deployed versions; verify the complete portal workflow with authorised test data; fix data-validation and duplicate-write issues; replace hard-coded admin access and plaintext password storage; enforce document access, retention and deletion controls. These are implementation work items, not claimed completed safeguards.

ICAI format basis

The public use-case library and a published Income Tax Calculator example support the problem statement, solution, role-of-AI and impact structure used here. The exact submission form, mandatory fields, category options and upload limits were not visible without SSP-mobile sign-in. This is not a verified hackathon template.

Source record

ICAI use-case library: https://ai.icai.org/usecases.php

Published example: https://ai.icai.org/usecases_details.php?id=138

ICAI sign-in: https://ai.icai.org/registration.php

Private code repository: https://github.com/cashalinisomani/gst-reg

Google Antigravity (tool description): https://antigravity.google/

Read directly on 14 October 2026: frontend/index.html; frontend/package.json; frontend/public/script.js; backend/package.json; backend/server.js; backend/database.js; automation/package.json; automation/runner.js; Dockerfile. Repository last updated 5 October 2026. No live GST filing or production deployment was tested for this document.