The AI Assistant CPAs Can Actually Use on Client Work

Understand anything
on your screen.
Privately.

Highlight any text — a confusing client email, an IRS notice, a dense contract clause — and ask Eduardo to explain it. Speak or type. Get a plain-English answer in seconds. Nothing uploaded. Nothing stored. Nothing leaves your computer.

100% local AI — no internet required
Works on Mac + Windows
One-time payment, no subscription
🍎 macOS🪟 Windows
Hey Eduardo
📋 Selected Text
“...the amended Schedule K-1 reflects an additional $47,200 in passive income attributable to the revised partnership allocation as of Q3...”
What does this mean for my client's tax bill?
Eduardo
The $47,200 flows to Schedule E on their 1040. At a 32% bracket that's roughly $15,100 additional federal tax. Worth checking if they have passive losses from prior years that could offset...
Ask anything about your context...
🔒 Processing locally — zero data uploaded
653,408
licensed CPAs in the US
46%
of workers upload sensitive data to public AI (KPMG 2025)
40M+
US adults process information differently
0 bytes
of your data ever leave your machine

Your team is already using ChatGPT on client files.

They're not doing it to cause trouble. They're doing it because it works. But every paste is a liability you can't walk back.

⚠️

Client data is landing on OpenAI's servers

A 2025 KPMG survey found 46% of workers admit uploading sensitive company information to public AI tools. At CPA firms that means SSNs, tax returns, and financial statements — your clients' most private information sitting on someone else's servers.

⚖️

Your duty of confidentiality has no cloud exception

State CPA ethics rules are clear — client financial data must be protected. A February 2026 federal court ruling confirmed that public AI chat tools are not a confidential channel. Anything your staff types in can be stored, processed, and subpoenaed.

🔁

The copy-paste workflow is costing you hours

Switching between client documents and a browser AI tab dozens of times a day isn't a strategy — it's friction. Eduardo lives next to your work. Highlight the confusing part, ask your question, get an answer, keep moving.

Four steps. Fifteen seconds. Total privacy.

Eduardo lives quietly in your menu bar and activates the moment you need it — on any app, any document, any screen.

1

Highlight the confusing part

Select any text on your screen — a paragraph from a client email, a line on an IRS notice, a clause in a document. Just the snippet that isn't clicking.

2

Hit your hotkey

Press ⌘+Shift+Space on Mac or Ctrl+Shift+Space on Windows. Eduardo's panel appears with your selected text already loaded as context.

3

Ask your question

Type or speak your question in plain English. "What does this mean?" "What should I watch out for?" "How do I respond to this?" Eduardo answers in seconds using AI running entirely on your machine.

4

Keep the conversation going

Follow up. Go deeper. Think out loud. Eduardo holds the full context of your session — like having a brilliant colleague sitting next to you who never forgets what you were just discussing.

Everything you need. Nothing you don't.

Built for professionals who work with sensitive information every day — not for casual chatting.

🖥️

Works on any screen, any app

Email, PDFs, Word docs, spreadsheets, browser tabs, Zoom transcripts. If it's on your screen and you can highlight it, Eduardo can help you understand it.

🎙️

Talk to it — don't just type

Hold the mic button and speak your question out loud. Fast, accurate voice-to-text runs completely offline. No typing required.

🔊

Reads answers back to you

Turn on voice output and Eduardo speaks its response aloud. Hands-free AI assistance while your eyes stay on the document.

🧠

Local AI that actually keeps up

Powered by Qwen3 9B — a best-in-class language model running entirely on your hardware. Excellent at comprehension, explanation, and analysis of what's on your screen right now.

☁️

Two modes. You control which one.

Local mode is the default — for anything touching client data. Nothing leaves your machine. Cloud mode is opt-in — connect your own API key for general research questions with no client data involved. The app always shows which mode is active.

💬

Full conversation memory per session

Eduardo remembers everything within a session. Ask a follow-up ten minutes later and it knows exactly what you were discussing. No re-pasting context, no starting over.

Your client data is too valuable to trust to anyone else.

Hey Eduardo was built on one principle: your data belongs to you. Not to OpenAI. Not to Anthropic. Not to us. It never leaves your computer — full stop.

This isn't a privacy policy. It's the technical architecture. There's no server to breach because there is no server. There's no cloud account to compromise because there is no cloud. Eduardo runs on your machine. Your conversations stay on your machine. That's the whole design.

🔒

100% on-device processing

All AI inference runs on your CPU or GPU. No internet connection required after the initial model download. The model lives on your hard drive.

🚫

Zero telemetry, zero tracking

We don't collect usage data, crash reports, or anything else. Nothing phones home. We genuinely cannot see what you're doing — by design.

🗑️

Sessions cleared on close

Conversation history lives in memory only. Close Eduardo and it's gone. No log files, no history stored on disk, no residue.

Client confidentiality preserved

Use AI on client documents with zero ethical exposure. What stays on your machine stays between you and your client — exactly as your professional obligations require.

Some people read a document and immediately get it. Others need to talk it through.

If you've ever read a paragraph three times and still couldn't tell someone what it means — Eduardo was built for you.

Not because anything is wrong with how you think. But because dense written language wasn't designed for the way a lot of brilliant people actually process information.

Highlight the part that isn't clicking. Ask Eduardo to explain it in plain English. Speak your question if you'd rather not type. Get an answer you can actually use.

That's it. No complicated setup. No learning curve. Just understanding.

40M+
US adults have dyslexia — most undiagnosed
35%
of US entrepreneurs are dyslexic (Cass Business School)
25%
more satisfied with AI tools than neurotypical users (UK Dept for Business & Trade, 2025)
“I've white-knuckled my way through the business world. These tools help so much.”
— Senior director, diagnosed with ADHD as an adult
“If even one client found out I was pasting their tax return into ChatGPT, it would end my practice. I needed AI that was completely private. Hey Eduardo is exactly that.”
Principal, 6-person CPA firm — Denver, CO
Placeholder quote — will be replaced with verified customer testimonial post-launch.

Pay once. Use forever.

No subscriptions. No per-user fees. No monthly surprises. Buy Eduardo once and it's yours — including all updates for the period specified.

Pro
$49
One-time · 1 year of updates included
Everything you need to use AI confidently on client work, with your data staying exactly where it belongs.
Full local AI (Qwen3 9B, runs on your device)
Unlimited sessions and conversations
Voice input and voice output
Mac + Windows — one license covers both
Connect your own API key (Anthropic, OpenAI, OpenRouter)
Conversation export (.txt or .md)
1 year of free updates
Business
$149
Per seat · One-time · All future updates
For firms deploying Eduardo across a team. Includes a commercial license, formal invoicing, and priority support.
Everything in Lifetime
Commercial use license
Formal invoice for expense reporting
Priority support (same-day response)
Team onboarding guide and documentation
Volume discounts at 5+ seats
Dedicated account contact

Prices locked in for early access members. Be first to know when we launch.

All plans include a 14-day money-back guarantee, no questions asked. · Need 10+ seats? Contact us at hello@heyeduardo.app

Common questions.

Yes — completely. The AI model is downloaded once during setup and runs entirely on your CPU or GPU after that. You can use Eduardo in airplane mode, in a secure client facility, anywhere with no connection. The only time any network activity occurs is if you deliberately choose to connect your own cloud API key for general research.
Hey Eduardo works on Mac (Apple Silicon or Intel) and Windows. We recommend at least 8GB of RAM. On Apple Silicon the AI runs via Metal GPU acceleration and is noticeably fast. On Windows with an NVIDIA GPU it uses CUDA acceleration. CPU-only mode works on all machines but is somewhat slower for generating responses.
Local mode is the default — everything runs on your machine and nothing is sent anywhere. Use this for anything involving client data. Cloud mode is opt-in — you add your own Anthropic or OpenAI API key and Eduardo can reach the internet for general research questions that don't involve client information. The app always clearly shows which mode is active so you always know.
The local AI model is excellent at comprehension, explanation, and analysis — but it won't know about recent legislative updates. Eduardo is a thinking partner, not a tax research tool. Use your existing Thomson Reuters or Bloomberg Tax subscription for current law questions, and use Eduardo to help you understand, explain, and work through what you're reading. For general research questions with no client data involved, cloud mode gives you access to current information.
Eduardo works best on snippets — a paragraph, a clause, a specific line item — rather than entire documents at once. This actually fits how most professionals use AI in practice: you highlight the confusing part and ask about that specifically. For full-document analysis workflows, the cloud mode with your own API key handles larger context windows.
When you use ChatGPT, everything you type goes to OpenAI's servers, is stored, and can potentially be used for model training. A February 2026 federal court ruling confirmed that public AI chat is not a confidential channel — anything typed in can be subpoenaed. Eduardo is the architectural opposite: your words, your documents, and your questions never leave your machine. There is no server on our end to breach because we never receive your data in the first place.
No free tier, but every plan includes a 14-day money-back guarantee — no questions asked. Download it, use it with your actual work for two weeks, and if it doesn't deliver clear value you'll get a full refund. We think that's a more honest test than an artificially limited demo version.

Your clients trust you with their most sensitive information. Honor that trust.

AI that respects client confidentiality isn't a luxury — it's what your profession demands, what your clients expect, and what your own instincts have been telling you. Eduardo makes the right choice the easy choice.