THE AI FOR LAZY PEOPLE
Too lazy to do it?Spark already did.
Spark remembers what matters, notices what you’re putting off, and gets things done before you even ask.
Your memories. Your schedule. Your digital life. Finally working together.
ONE PROMPT · A SWARM OF SPARKS
Say it once. Watch the swarm do it.
Spark breaks one request into dozens of small jobs (read a site, write the page, upload the files) and works them side by side.
An illustration of how Spark splits up work. In the app, Spark runs a few browser agents at once; here the swarm is multiplied and sped up.
A day without doing anything
Your life keeps moving. Even when you don’t.
You have better things to do than manage every little thing. Spark handles the busywork.
You
Spark
You: Still in bed.
Checked your inbox and found an important deadline.
You: Second coffee.
Prepared the application you kept postponing.
You: Lunch, phone face-down.
Found an event related to your interests.
You: A long walk.
Resolved a scheduling conflict.
You: Dinner. No laptop.
Prepared tomorrow’s priorities.
Less managing. More living.
Concept demo An illustrative day, not a recording.
Superlong memory
It remembers. Even when you don’t.
That random idea from three months ago? Spark knows when it matters again.
Spark keeps the meaningful parts of your conversations — plans, preferences, people, dates — in a memory you can open, edit and delete. Next, it learns to bring them back at exactly the right moment.
MARCH 12
Spark records the intention.
JUNE 24
Spark connects it to the winter trip.
OCTOBER 15
Concept demo Illustrative conversation.
Your anniversary is coming up.
That internship you wanted? Applications are open.
You mentioned wanting to start exercising again.
Your planned trip is getting closer.
Not just memory. Memory with initiative.
Spark doesn’t just store old messages. It understands when they become relevant again.
The memory universe
Everything you’ve mentioned. Quietly connected.
Scattered memories drift until they meet. Then they become one next step.
Concept demo A visualization of how Spark links context. Illustrative.
Spark takes the initiative
You shouldn’t have to ask twice. Or even once.
Spark notices what matters, recognizes when you’re falling behind, and prepares the next step.
Traditional AI
Remind me to submit my application.
Sure, I’ll remind you.
Waiting for you to come back…
Spark Concept demo
Nobody askedThat application closes tomorrow. I’ve prepared everything using your previous documents. Ready to submit?
Spark prepares. You press submit.
Detects relevant deadlines, upcoming events and unfinished intentions from the sources you’ve connected.
Decides when an old intention deserves attention, based on timing, context and urgency.
Prepares or completes the workflows you’ve authorized, without making you repeat yourself.
From “I should” to “It’s done.”
From memory to action
Remembering is easy. Getting it done is the magic.
One thread runs from a passing remark to a finished task — and you can see every step.
Concept demo Illustrative flow. Today Spark acts on what you ask and on the routines you set; acting on old intentions is in development.
One conversation can become an action months later.
Give Spark permission to handle it
Give it permission. Get your time back.
You’re in control of what Spark can handle on its own. Try it — every switch below works.
Prepare. Spark researches, drafts and prepares tasks for your approval.
- Calendar management
- Drafting emails
- Sending routine messages
- Event applications
- Browser workflows
- PaymentsAlways asks you
- Sharing sensitive infoAlways asks you
Send “Running 10 minutes late” to Mina?
Your 6:30 PM dinner — traffic on your route.- Moved “Design sync” to 3:30 PM
- Drafted a reply to Jordan
- Applied for a Design Week pass
- Stopped before checkout
This panel is a product vision: you can play with it here, but these exact settings aren’t in the app yet. What Spark does today: payments, permanent deletes and password changes only happen when you asked for them in your own words, and messages confirm before sending unless your request already said to send.
Your AI. Your rules.
Not just another AI that talks. One that actually does things.
Spark sees the screen the way you do — then moves the cursor, types into the field and checks the result before it moves on.
- Clicks.
- Types.
- Drags.
- Reads.
- Thinks.
- Fills forms.
- Scrolls.
- Calculates.
Not a mockup
Less explaining. More doing.
Sorting products on a shopping site. Finding assignments on a school portal. Unedited recordings of Spark at work.
Real product recordingsThe film
Meet Spark.
Fifty-seven seconds of what a day looks like when your schedule runs itself.
Measured, not claimed
Same task. Less waiting.
We ran Spark and OpenAI Codex on the same live sites and apps, with the same model, and timed every run.
1.6× faster computer use than Codex
Timed on the same live sites and apps with the same model. Spark looks once and runs several actions in one go.
It doesn’t take your mouse
Websites run in Spark’s own background tab, Windows apps through window messages, so your mouse and keyboard stay yours while it works. (A few apps borrow the screen only while you’re idle.)
KakaoTalk in 27.2 seconds
Sends a KakaoTalk message on Windows in 27.2 s. Codex took 69.2 s for the same task.
- Spark
- Codex
Kakao Map · route Gangnam → Seongsu
Kakao Map · a restaurant’s phone number
Danawa · lowest price
Wikipedia · look up a station
CGV · tonight’s showtimes
KakaoTalk app · send a message (desktop)
Upload a file to a web form
Measured Oct 19–20, 2026 on live sites and the Windows KakaoTalk app; one run per task; both on gpt-6-astra. Codex times exclude its approval-review pauses. Results vary with sites, network and model load.
19 tools with a defined connection path
OAuth 9 · paste a token 5 · through the browser 5
GmailOAuth
Google CalendarOAuth
SlackOAuth
NotionOAuth
DiscordOAuth
NAVER WORKSOAuth
OutlookOAuth
GitHubpaste a token
Linearpaste a token
Jirapaste a token
Asanapaste a token
Trellopaste a token
Zoomthrough the browser
Microsoft TeamsOAuth
Dropboxthrough the browser
Google DriveOAuth
Salesforcethrough the browser
HubSpotthrough the browser
Zendeskthrough the browserWhere an OAuth app is not approved yet, it connects by token or through the browser. Sites that are not on the list can still be handled through the browser.
Not a bot in the cloud —
your own Chrome does the work.
Spark attaches to the Chrome profile on this machine and works already signed in. Instead of shipping screenshots it reads the page structure, and calls the model only to decide.
An example of the events it streams
{"type":"step","action":"click",
"timing":{"perceive_ms":…,
"decide_ms":…,"tier":"t1"}}Your Chrome profile, as is
No second login. It uses the session you already have open.
Page structure, not screenshots
It reads the elements and text in view to pick the next action, and looks at the pixels only when there is nothing to read.
It leaves evidence
When it extracts a value or finishes, it saves that screen and sends it back with the result.
Set it once,
and it keeps going with the app closed.
Every 1 minute while this PC is on, and every 5 minutes from the cloud when it is off. Results come back to the chat that created them.
When
What
Where to
Even with the app fully closed, the server pushes deadline reminders and a morning digest every hour.
Check the university portal
Every morning at 9, pull only the startup and scholarship items from my school portal
Routines running today
Sample — not real account dataThe list below is a sample used to show the layout. Only the times come from your own clock.
- --:--University portal noticesOnly the items matching your keywordsDone
- --:--Store ordersNew orders and unanswered questions, as a tableDone
- --:--Three competitors’ pricesComparing only what changedRunning
- --:--Contest deadlinesOnly what closes within two weeksQueued
Privacy & trust
Personal enough to help. Private enough to trust.
Spark remembers so you don’t have to — and shows you what it keeps, where it runs and what it did.
Payments, permanent deletes and password changes only run when you asked for them in your own words.
stop_before_paymentGmail, Calendar, Notion, Slack and the rest are separate, opt-in connections you can remove whenever you like.
See every memory Spark keeps. Edit, pin or delete any of it — or switch memory off and wipe it, server copy included.
Every answer comes with a step-by-step record of what Spark opened, clicked and typed, with secrets redacted.
redactMessages confirm before they go unless you told Spark to send them. Logins and 2FA codes stop for you or your vault.
wait_for_2fa_codeMemory recall and browser work run on your PC. To think, Spark sends your request to AI models (OpenAI, Google Gemini). Memories, chats and tasks sync to your Spark account so they follow you between devices — they don’t stay only on your device, and they are never used to train models.
Saved logins are protected by Windows DPAPI or the macOS Keychain on your computer and stored with AES-256-GCM in your account. Passwords never go into memory.
AES-256-GCMStart lazy.
Stay lazy.
Start free. Upgrade when Spark is doing more of your busywork. No hidden costs, ever.
Memory is on every plan. Autopilot and the permission controls shown above are in development.
Free
Meet your new AI sidekick.
- AI memo analysis (50/day)
- AI chat (100 messages/day)
- Web search (15/day)
- 10 memo pages
- Auto task classification
- Web dashboard
Pro
More done. Less busywork.
- AI memo analysis (unlimited)
- AI agent web search
- Google Calendar sync
- Gmail integration
- Unlimited memo pages
- Web & mobile sync
- Auto URL analysis
- Priority support
Ultra
Your routines, on autopilot — 24/7.
- Everything in Pro, plus:
- 24/7 autonomous AI agent
- Round-the-clock schedule processing
- Executes tasks while you sleep
- Auto-reschedules & resolves conflicts
- Non-stop Gmail / Notion / Calendar monitoring
- Browser use — it opens, reads and summarizes sites itself
- Priority agent compute
Team
Enterprise plan for team collaboration
- All Pro features included
- Shared team memo board
- Team task assignment
- Real-time collaborative editing
- Unified team calendar
- API access
- Custom AI prompts
- Dedicated support
Frequently asked questions
What does Spark do?
It is a scheduling browser agent: at the time you set, it opens sites in your browser, reads them, and brings back a summary. Routines repeat hourly, daily, weekly or on a custom interval, and results return to the chat that created them.
Do routines run with the app closed?
While this PC is on, Spark checks every 1 minute. With the app closed, the cloud checks every 5 minutes and runs them instead. The cloud does not have your Chrome logins, though — anything behind a login works properly when this PC is on.
Will it pay or verify for me?
No. It stops before the final payment button and asks you for two-factor codes. When it hits a bot wall it hands the screen over and waits.
Where are my passwords kept?
Browser sessions stay in the Chrome profile on this PC. Credentials you save to the vault are encrypted by the server with AES-256-GCM and returned only to an authenticated request from you. Password-shaped values are scrubbed from the run log.
What does it connect to?
Gmail, Google Calendar, Google Drive, Slack, Notion, Discord, NAVER WORKS, Outlook and Microsoft Teams connect by OAuth; GitHub, Linear, Jira, Asana and Trello by pasting a token; Zoom, Dropbox, Salesforce, HubSpot and Zendesk through the browser. Sites that are not on the list can still be handled through the browser.
Is there a free plan?
Free gives you 50 AI analyses, 100 chat messages and 15 web searches a day, plus 10 memo pages. Beyond that, Pro is $19.99/month, and Ultra — for autonomous runs that continue with the app closed — is $49.99/month.
Spark isn’t another thing to manage.
It’s the thing that manages the rest.
Join the waitlist and we’ll tell you the day Spark opens to everyone.
Get Spark