AI-generated conceptual editorial illustration. Not a documentary photograph.

Agentic AI

The briefing

Google's agent kit reaches 1.0 in Kotlin, with on-device pieces

By Ashlr.AIAI-assisted briefingPublished Sep 20, 2026, 3:00 PM UTC3 min read

Google's Agent Development Kit reached 1.0 for Kotlin, adding Android and JVM targets. InfoQ reviewed it on September 20; Google's own post is dated September 9, and one on-device path is still beta.

In this story

Browse all stories

Agentic AI

The briefing

Perplexity Teammates enable persistent multi-agent workflows

AI-generated illustration

Nov 9, 20262 min read

Perplexity launched Teammates, a system that lets multiple agents coordinate tasks, maintain shared project state across sessions, and continue working after a user closes the browser tab. The feature is available now to web users on the Max plan.

What changed

On September 9, 2026, Google released version 1.0 of its Agent Development Kit (ADK) for Kotlin. The kit is for building AI agents in Kotlin, Java and Android apps. InfoQ wrote about it on September 20. Google says the Kotlin version now matches ADK 1.0 for Python and Java, and adds extras for Android.

Google lists what 1.0 covers. Agents can hand work to other agents. Long chats can be summarized to stay within limits. An agent can pause and ask a person to confirm a sensitive action. Sessions can be saved and restored. Java apps can call the agents directly. Tool definitions are generated when you build the app, so a mistake shows up at build time.

The on-device part is not one thing. Google lists LiteRT-LM and ML Kit for running models on the phone, and labels ML Kit as beta. Cloud work goes through Firebase AI Logic or Vertex AI. Google\'s Android example uses Gemini 3.8 Flash through Firebase AI.

Google calls the release production-ready. That is Google\'s word, not an outside test. Google\'s own banking example says it is for demonstration only and is not designed to meet any compliance requirements.

What it means for your work

For builders

If your agents run in a Python service that your phone or Java app calls over the network, ADK for Kotlin lets you move them into the app. That removes one network hop. Try it with one small agent first. Test the things Android is hard on: whether a session survives when the system closes your app, what happens on a phone where the on-device model is missing, and whether you can still swap the model provider. Turn on user confirmation for any action that moves money or changes data.

For business

Ask where each feature actually runs. A model running on the phone keeps the text off a vendor\'s servers, which can matter for a privacy review. But only the parts that truly run on the phone get that benefit, and Google labels the ML Kit route beta. Anything that goes through Firebase AI Logic or Vertex AI leaves the device and is billed and processed under the cloud terms. Have your team say which path each feature uses before anyone tells a customer it runs locally.

Put it to work

Pick one agent feature. List its tool calls and mark which ones could finish with no network. For each, write down which model path it uses and whether Google lists that part as stable or beta. The open question is not whether Kotlin works. It is how much of your agent actually runs on the phone. That list tells you whether an on-device pilot is worth the time.

Go to the source

Source material behind this briefing. Our practical implications are Ashlr’s interpretation.

Announcing ADK for Kotlin 1.0: Building Production-Ready AI Agents in Kotlin, Android, and Beyond (opens in a new tab)
Google Developers BlogSep 9, 2026
Google Agent Development Kit for Kotlin Reaches Feature Parity with Python, Supports On-Device AI (opens in a new tab)
InfoQSep 20, 2026

This briefing was prepared with AI assistance from the linked sources. The analysis may include interpretation; consult the original sources for complete context. Cover artwork is illustrative.

Editorial approach Suggest a correction

Agentic AI

Ashlr release

Phantom fleet board and mobile parity updates

AI-generated illustration

Nov 7, 20262 min read

Merged bridge jobs added a live resource board to Fleet and mobile parity for sessions and approvals in Phantom. Ashlr Stack gained worktree cleanup reports and command updates.

Agentic AI

Ashlr release

Phantom fleet verbs and product guides added

AI-generated illustration

Nov 1, 20262 min read

Phantom gained agent-native fleet commands for lane targets, reassigns, and model sets. Multiple products received new documentation and regression tests.

Google's agent kit reaches 1.0 in Kotlin, with on-device pieces

Keep exploring

Perplexity Teammates enable persistent multi-agent workflows

What changed

What it means for your work

For builders

For business

Put it to work

Go to the source

Phantom fleet board and mobile parity updates

Phantom fleet verbs and product guides added