A Singapore startup just put an AI agent where you spend most of your screen time: the keyboard. On Tuesday, Acti (short for “action”) launched an agentic keyboard for iOS and Android, and according to TechCrunch AI, it doesn’t just predict your next word. It can take actions for you, pulling AI tools straight into the apps you already use for email, messaging, and social media.
Founder and CEO Young Wang frames the pitch around a problem most of us know well. You’re mid-conversation, you need an AI’s help, and you have to jump out to a separate chatbot app, then come back. Acti wants to kill that switch. “Today’s AI agents are fundamentally limited because user context stays fragmented across separate apps,” Wang told TechCrunch AI. Because the keyboard “sits across all of them,” he argues, Acti can build a context layer that belongs to the user rather than any single platform.
What it actually does
The idea is simple once you see it in action. If a friend asks where to eat, Acti can drop a local recommendation right into the chat. Mention a stock, and it can pull the live price into the same thread. No tab-switching, no copy-paste.
Key features TechCrunch AI highlights:
- Skills: custom shortcuts tied to a single key. Long-press “T” to translate a message, or hit “C” to fire off a meeting link. You program a key once, and it runs a multistep task automatically.
- Plain-language building: you don’t need to code. Describe what you want, and Acti builds the Skill for you.
- A Skills marketplace: Skills can stay private or get shared publicly. Early testers have already published ones for real-time World Cup data and Polymarket links.
- Local-first privacy: personal context stays on your device by default. Acti says it doesn’t access or store private messages unless you invoke a feature that needs external processing.
Under the hood, Acti runs on Google’s Gemini models. Wang picked them for the balance of intelligence, speed, multilingual performance, and cost. Gemini also powers the Skills engine that turns a plain-language request into a working shortcut.
Why this founder, why now
Wang isn’t new to keyboards. He spent a decade at Baidu growing its Facemoji Keyboard past 300 million daily active users. What changed his thinking was the arrival of large language models. “Text was no longer just something people typed; it had become a carrier of intent,” he said. “And in many everyday contexts, that intent can now be directly translated into action.”
That’s the real bet here. The keyboard is one of the most universal surfaces in computing, and Acti wants to rebuild it as the entry point for AI actions instead of a passive text field.
Traction and funding
The early numbers suggest people get it. Ahead of launch, testers built over 1,000 Skills in less than two weeks. Acti also told TechCrunch AI it closed $5.3 million in seed funding, led by BITKRAFT Ventures. “We backed Acti because this team has a real shot at owning the next phase of human-computer interaction,” said partner Jonathan Huang.
The team carries more Baidu pedigree. CTO Mike Sun was the founding technical lead on Yike Album, Baidu’s cloud-photo platform that scaled past 10 million daily active users. CSO Junbo Yang joined from HashKey Capital.
Availability and the caveats
Acti is live now on both iOS and Android. The business model is still forming. The company plans subscriptions for more advanced models, higher daily usage limits, and premium features, and the Skills Hub could add monetization down the line.
What stands out to me is the positioning. Most consumer AI plays ask you to adopt a new app and change your habits. Acti flips that by embedding itself in a surface you already touch thousands of times a day. The open question is trust. A keyboard sees everything you type, so the local-first promise isn’t a nice-to-have, it’s the whole ballgame. If Acti holds that line while proving the actions are genuinely useful, it has a shot at owning a piece of real estate no chatbot can reach.
For the full breakdown, including the launch demos, check the original report from TechCrunch AI.