A privacy-first, single-file HTML tool that turns a trial balance, P&L, or balance sheet into a complete, risk-stratified, assertion-mapped audit checklist — in under 30 seconds, using Claude AI.

No installation. No server. No subscription. Download one file, open it in your browser, and start using it.

Built by Vijay, a Cost Accountant and Government of India Certified GST Practitioner, for Indian CA firms, cost accountants, and audit trainees.

→ Visit the site · → Launch the tool directly · → Docs & FAQ

This repo is a small three-page site, not just a single file:

Only ai-audit-checklist.html talks to the Anthropic API. index.html and docs.html are static marketing/documentation pages with no API calls and no CSP requirement of their own.

Building an audit checklist manually means receiving the client's financials, reviewing them line by line, digging out the SA (Standards on Auditing) manual to map assertions, drafting procedures from last year's Word template, and printing it out for the team. It typically takes 6 to 10 hours — and the quality depends entirely on who's building it. A senior CA with 15 years of experience catches a red flag in the debtors' aging that a junior staff member misses entirely. None of that intelligence carries forward systematically; it lives in someone's head, not in the checklist.

This tool solves the time problem, the consistency problem, and the "tribal knowledge" problem at once.

1. Engagement details — entity name, financial year, entity type (Private Limited, Public Limited, Partnership, LLP, Proprietorship, Trust/NGO), and audit type (Statutory Audit, Internal Audit, Tax Audit u/s 44AB, Limited Review, Due Diligence).
2. Focus area chips (optional) — tap Revenue Recognition, Inventory, Related Party Transactions, Fixed Assets, and more, to guide the AI's attention. The AI identifies risk areas from the data itself even without any chips selected.
3. Paste financial data — trial balance, P&L, or balance sheet. Doesn't need to be perfectly formatted; paste whatever you have.
4. Click Generate — within ~30 seconds, Claude reads every figure, scores every area, maps every SA assertion, and drafts every procedure.
5. Work through the checklist on screen — tick off items, watch the progress counter, and copy everything to your working paper in one click.

For CA practice, client confidentiality isn't optional — so here's exactly how this tool protects it, technically, not just as a promise:

• 100% client-side. All logic runs inside your browser, on your machine. There is no server, no backend, no database. Nothing runs outside your computer.
• CSP-enforced isolation. A Content Security Policy meta tag at the top of the HTML file instructs your browser — before any JavaScript runs — to block every outbound connection except to api.anthropic.com. Even if malicious code were somehow injected into the page, it cannot send your data anywhere else. The browser itself refuses the connection.
• Session-only API key. Your Anthropic API key lives only in sessionStorage, wiped automatically the moment you close the tab. It is never written to disk, never written to a cookie.
• Auto-clear inputs. After the AI responds, the financial data you pasted is automatically wiped from the form (toggle available in the UI).
• Anthropic zero retention. Anthropic's API does not store, log, or train on data submitted via API calls — each call is stateless and ephemeral.
• Offline capable. The file works without an internet connection except for the moment you click Generate, and even then it only ever reaches api.anthropic.com.

You can open the HTML file in any text editor and see the exact CSP line enforcing this — nothing here is a black box.

• Any modern browser (Chrome, Edge, Firefox, Safari).
• Your own Anthropic API key — get one free at console.anthropic.com. Takes about 5 minutes; a few hundred rupees of credit covers a full year of use for most firms (see Cost & Impact below).

No coding knowledge, installation, or IT support needed.

You can either browse to the tool from the site/GitHub Pages URL, or download ai-audit-checklist.html on its own — it's fully self-contained and works standalone either way:

1. Download ai-audit-checklist.html from this repo (green "Code" button → Download ZIP, or right-click → Save Link As on the raw file).
2. Open it by double-clicking — it opens directly in your default browser. No server, no build step.
3. Paste your Anthropic API key into the API Key field at the top (starts with sk-ant-). It's used only for this session and never leaves your browser except to reach Anthropic.
4. Fill in engagement details — entity name, FY, entity type, audit type. Optionally list your team's names (comma-separated) if you want to assign checklist items later.
5. Optionally select focus area chips for the areas you want the AI to pay extra attention to.
6. Provide the financial data — either paste a trial balance/P&L/balance sheet as text, or switch to the "Upload PDF" tab and drag one in directly. Optionally toggle on Prior-Year Comparison and paste last year's figures to get automatic year-on-year risk flags.
7. Click "Generate Audit Checklist." Within seconds, a risk-stratified, assertion-mapped checklist appears on the right.
8. Work through it — click items to tick them off, watch your progress counter, expand/collapse sections as needed.
9. Click "Copy All" to paste the full checklist into your working papers.
10. Click "Clear Session" when done, or simply close the tab — the API key and any leftover data are wiped automatically.

There is no hosting cost, no software licence, and no maintenance fee — just the pay-as-you-go Anthropic API cost for each checklist generated. Consistency is the less obvious but equally important benefit: a junior CA using this tool produces the same assertion-mapped, risk-scored output as a senior CA would, freeing the firm's most experienced people for judgment calls rather than checklist drafting.

Is the AI output always accurate? No — and it isn't meant to replace professional judgment. The AI produces a first draft. It reduces preparation time from hours to minutes, not to zero. The CA reviewing the output remains fully responsible for the final checklist and the audit opinion built on it.

What about client confidentiality under ICAI guidelines? This tool is a productivity aid, exactly like a Word template or a calculator — the CA still owns the judgment and the output. The CSP enforcement and session-only architecture ensure no client data is stored anywhere or shared with anyone beyond the single API call to Anthropic needed to generate the checklist.

Can it handle different types of entities? Yes — Private Limited, Public Limited, Partnership, LLP, Proprietorship, and Trust/NGO are all supported, and the AI adjusts its procedures and compliance flags based on the entity type selected.

What if the client's data is in Tamil, Hindi, or another regional language? Claude understands multiple Indian languages. You can paste data or notes in Tamil, Hindi, or another regional language and the AI will process it correctly.

What does it cost per year? Roughly ₹200–500 annually for a firm running 40 audit engagements, or ₹500–1,200 for 100 engagements — all of it the pay-as-you-go Anthropic API cost. There is no subscription or licence fee for the tool itself.

This tool generates a first draft of an audit checklist using AI. It does not constitute audit evidence, a completed audit programme, or professional advice on its own. The Chartered Accountant or audit professional using this tool remains fully responsible for reviewing, verifying, and exercising professional judgment on every item before it is relied upon in any engagement. Always cross-check AI-generated red flags and procedures against the actual client data, applicable SA standards, and your firm's quality control policies.

You can host this tool online — for yourself or your firm — using GitHub Pages, entirely free:

1. In your repo, go to Settings → Pages.
2. Under Branch, select main and / (root), then Save.
3. GitHub will publish your repo at (https://vijay301299.github.io/ai-audit-checklist-generator/).
4. Because the repo already has an index.html at the root, that URL opens the landing page automatically. The tool itself lives at (https://vijay301299.github.io/ai-audit-checklist-generator/), and docs at .../docs.html — both are linked from the landing page nav, so you only need to share the one root URL.

Important: hosting the files on GitHub Pages does not weaken the privacy guarantees above. The Content Security Policy in ai-audit-checklist.html is embedded in that file itself and is enforced by the visitor's own browser, regardless of where it's served from. Whether someone opens the tool from their Desktop or from a GitHub Pages URL, the browser still blocks every outbound connection except to api.anthropic.com. index.html and docs.html are plain static pages with no API calls at all — GitHub Pages is only a convenient way to distribute the same files, not a server that touches your data.

Phase 1 & 2 — Delivered. Trial balance / P&L / balance sheet input via paste or PDF upload (read natively by Claude — no OCR step needed), risk-stratified checklist generation, SA assertion mapping, data-specific red flags, prior-year comparison with automatic year-on-year movement flagging, team assignment (assign checklist items to staff, included in the Copy All output), interactive tick-off, and copy-to-working-papers. All of it is live and working today.

Phase 3 — Scaling to a product.

• White-label version for CA firms, with custom branding and a multi-client dashboard.
• Industry-specific checklist templates — Manufacturing, NBFC, NGO, Real Estate — each with domain-specific procedures pre-loaded.
• Direct integration with Tally and Zoho Books exports, so financial data flows in automatically.

The foundation for Phase 3 already exists in this repo: same single HTML file, same privacy model, same zero-cost deployment.

MIT — see LICENSE. Free to use, modify, and white-label for your own firm.

Built by Vijay — Cost Accountant at L.G. Balakrishnan & Bros. Ltd. and Government of India Certified GST Practitioner (Enrolment No. 332500009598GPC), running VJ GST & Tax Consultancy independently in Coimbatore, Tamil Nadu. First shown live at the IGNITE Audit Workshop Hackathon, July 2026.