Native macOS · Apple notarized
Tonttu gives AI agents structured control over your macOS apps and browser — using the UI itself, not just screenshots. Faster execution, fewer tokens, better results.
Demo coming soon
Approach
Most computer-use agents screenshot your display for every single action — burning thousands of vision tokens each time. Tonttu takes a smarter approach: it reads the actual UI structure first and exposes named actions your agent can call directly. Screenshots are there when needed, but most tasks never require them.
Screenshot-based agents
Tonttu
You don't need Claude Max for this
Claude Dispatch gives you computer control inside Cowork — but it's Claude-only, requires a Pro or Max plan, and relies on screenshots for every action. Tonttu works with any AI you already use, reads the UI structure first, and costs $39 once.
If you're already on a Max plan, Tonttu still saves you tokens on every task — structured tool calls instead of vision-heavy screenshot loops.
Works with any AI
Claude, GPT, Gemini, OpenClaw, local models — not locked to one provider.
No screenshot overhead
Structured text instead of vision tokens. Same task, fraction of the cost.
$39 once — no plan required
Bring whatever AI subscription you already have. Tonttu is the bridge.
Works with OpenClaw
OpenClaw agents gain structured access to every macOS app and browser on your machine — through named tool calls instead of relying on screenshots alone. Your existing skills and workflows stay the same. Tonttu just makes them faster and cheaper to run.
agent → screenshot display
agent → encode & send image to LLM
agent → receive coordinates
agent → click (x: 482, y: 316)
agent → screenshot again…
agent → click_submit_button
agent → fill_email("value": "...")
Demo recording
Describe what you want to show. Your AI walks through your product, performs every action live, and narrates each step with text-to-speech. You get an MP4 — ready to share on your landing page, pitch deck, or docs site. No screen recording, no editing, no retakes.
Tell your AI what to demonstrate. It navigates your app, clicks the right things, and shows the flow end to end.
Every step is narrated with natural text-to-speech. Choose the persona and tone — sales pitch, tutorial, QA walkthrough.
Get a polished video file. Drop it into your site, share it with investors, or post it — no editing software needed.
Sample recording coming soon
Capabilities
Any native app — spreadsheets, email, dev tools, design software. Your AI interacts through the accessibility layer.
Any website, any page. No extensions to install. Works through content security policies that block other tools.
AI-driven product walkthroughs with text-to-speech narration. Export MP4 directly.
Record a sequence of actions. Replay it on demand. Reusable workflows without writing scripts.
Claude, ChatGPT, Gemini, OpenClaw, local models — any agent that supports tool calls.
No telemetry. Everything runs locally on your Mac. Apple notarized and signed.
Pricing
The $39 pays for itself in saved tokens within a week.
One-time payment · Lifetime license
Requires macOS 14+ · Apple Silicon & Intel