How it works — ToDo's AI

You connect Gmail with one tap

Sign in with Google and grant read-only access to your inbox. That's the only permission we ask for — we can read emails, but we can't send, delete, or modify anything. Your Gmail account stays exactly as it is.

Read-only Gmail access — no sending, deleting, or modifying

Standard Google OAuth — we never see your Gmail password

Revoke access at any time from your Google account settings

Emails are fetched and filtered

We pull your recent emails and immediately discard anything that clearly isn't actionable — newsletters, automated receipts, marketing blasts, and notification emails that need no response. This keeps the AI focused on the emails that actually matter.

Bulk mailing infrastructure (unsubscribe headers) filtered out automatically

System notifications and zero-content automated messages skipped

Real emails from real people always make it through

The AI reads each email and classifies it

A local AI model reads the full email content and decides whether it contains something you need to do. It's not keyword matching — it actually understands context. An email saying "let me know if you're free Thursday" is classified as actionable. One saying "your order has shipped" is not.

Reads the full email body, not just the subject line

Understands implicit requests, not just explicit ones

Confidence threshold — low-confidence calls are skipped, not guessed

Tasks are extracted with full context

For emails that pass classification, the AI extracts a structured task: a clear title, due date (if mentioned), estimated time, priority score, and any sub-tasks. It writes these in plain language, not email language — so instead of "Re: see below regarding the attached form," you get "Sign and return the form for Dr. Martinez."

Clear, action-oriented task titles

Due dates parsed from natural language ("by end of week," "before Thursday")

Priority 1–10 based on urgency and sender context

Sub-tasks for multi-step requests

Tasks are categorized and deduplicated

Each task gets assigned to a category — Work, School, Personal, Finance, Health, and so on. If the same action comes up in multiple emails (two follow-ups about the same meeting), you get one task, not two. Your list stays clean.

Auto-assigned to categories based on content and sender

Duplicate detection via content hashing — no repeated tasks

Custom categories you can create in the app

The AI learns from your feedback

When you mark a task as confirmed or reject it as irrelevant, that signal goes back into the system. Over time, the AI learns your specific context — it figures out which senders always have actionable emails, which categories you care about most, and how you define "priority." The longer you use it, the more accurate it gets.

One-tap confirm or reject on each suggested task

Sender rules built automatically from your feedback patterns

Personalized fine-tuning once enough feedback is collected

Fetch

Pull emails via Gmail API

Filter

Remove bulk and noise

Classify

Actionable or not?

Extract

Title, date, priority

Save

Task in your app

Privacy first

Your data stays yours

The AI processing runs on our own servers — your emails are never sent to third-party AI providers.

Local AI

Email content is processed by a model running on our own infrastructure. Nothing is sent to OpenAI, Anthropic, or any third-party AI service.

Encrypted at rest

Your Gmail tokens are encrypted before storage. Even if someone got into the database, they couldn't use them.

Read-only access

The Gmail permission we request only allows reading. We cannot send emails, delete messages, or make any changes to your account.

From inbox to task list,without lifting a finger.