Google ADK for Kotlin 1.1: Android & JVM Agent Upgrade Guide
QUICK ANSWER
Google announced Agent Development Kit (ADK) for Kotlin 1.0 general availability on September 9, 2026, with full ADK 1.0 Core parity for Kotlin, Java and Android plus a Kotlin Multiplatform core, multi-agent orchestration, context compaction, human-in-the-loop confirmation, long-running tools, session resumability and Android-first extensions. The official google/adk-kotlin release feed now lists 1.1.0, released September 15, as the latest release. Version 1.1.0 adds Android AppFunctions guidance, an AppFunctionsToolset and conversion layer, plus fixes for workflow-agent session resumability, MCP request rejection behavior, runner event persistence and Vertex AI session IDs. For a new Kotlin agent project, evaluate the current 1.1.x release rather than copying a 1.0.0 launch snippet blindly. For an existing 1.0.x project, upgrade in a branch, test resumable workflows and tool calls, verify Android AppFunctions only where the app and platform support them, and keep MCP transport/configuration explicit. Google's 1.1.0 notes recommend Streamable HTTP over SSE for MCP connections. The 1.0 GA announcement also labels the ML Kit GenAI Android extension beta, so do not generalize the core GA status to every optional Android extension.
A Kotlin or Android agent project was started from the ADK for Kotlin 1.0 launch example and has not been reviewed against the newer 1.1.0 release
A resumable workflow agent completes or resumes incorrectly on 1.0.x and the team has not checked the 1.1.0 session-resumability fixes
An MCP tool call is rejected and the agent turn ends unexpectedly instead of continuing through the newer rejection-handling behavior
An Android app wants to expose supported AppFunctions to an ADK agent but lacks a bounded toolset and confirmation model
A team assumes every Android-first ADK extension is GA because the ADK for Kotlin core reached 1.0 general availability
Possible causes
The September 9 launch announcement describes the 1.0 GA baseline, while the official release feed moved to 1.1.0 on September 15
ADK for Kotlin spans core orchestration, Android-specific integrations, model providers, persistence and MCP connections, so component maturity and failure modes are not identical
Version 1.1.0 changed Android AppFunctions integration and fixed specific session, runner, Vertex AI and MCP behaviors that can matter to existing agents
Google's ADK documentation is updated independently of copied blog snippets and now provides an MCP documentation server plus machine-readable llms.txt indexes for current reference
HOW TO FIX IT
Work from the safest step to the harder repair.
Step 1. Inventory the ADK Kotlin artifacts and versions actually used by the app. Record core, processor, Android extension, model/persistence and MCP-related dependencies instead of assuming they all move in lockstep.
Step 2. Read the official 1.0 GA announcement for the architectural baseline, then compare the official 1.1.0 release notes before changing dependencies. Treat 1.1.0 as a release update, not as proof that every optional Android integration is production-ready.
Step 3. Create a reversible branch or checkpoint before upgrading. Keep the existing lockfile, Gradle version catalog and known-good build available so dependency or generated-code changes can be rolled back cleanly.
Step 4. Upgrade the ADK Kotlin packages through the project's normal dependency-management path and run a clean compile. Do not copy the 1.0.0 version numbers from the launch article into a project that intentionally tracks current releases.
Step 5. If the project uses resumable workflow agents, build a regression that pauses, serializes, restores and completes the workflow. The 1.1.0 release specifically fixes completion and resume behavior for workflow agents.
Step 6. If the project uses MCP, test both successful and rejected tool requests. Version 1.1.0 fixes behavior when an MCP server rejects a request, and its documentation recommends Streamable HTTP over SSE for MCP connections. Do not change transport solely by string replacement; confirm the server and client configuration support the selected transport.
Step 7. If the Android app uses AppFunctions, expose only the functions the agent genuinely needs. Review generated or converted argument types, authorization boundaries and user-confirmation requirements before allowing state-changing app actions.
Step 8. Keep human-in-the-loop confirmation around sensitive actions such as purchases, messages, destructive changes, account operations or privileged device/app functions. Framework support for confirmation does not decide your product's authorization policy for you.
Step 9. Re-run unit tests, integration tests and representative Android instrumentation/emulator flows. Include multi-turn context, tool errors, process restart/session restoration and AppFunctions paths that changed during the upgrade.
Step 10. Review persistence and privacy behavior separately for Room, AppSearch, cloud session services and model-provider traffic. Do not assume on-device persistence means model inference or every tool call is also on-device.
Step 11. For documentation-sensitive agent tooling, use Google's current ADK docs rather than a frozen prompt corpus. The ADK docs publish llms.txt and llms-full.txt indexes and document an MCP-based docs server for supported coding tools.
Step 12. Record the verified ADK version, test results, optional-extension maturity and rollback point in the release note so a later framework update can be evaluated against a known baseline.
Need the actual code? Go to GenesisCodeDoctor.com to search the Code Store or request code for the exact platform, error, and repair you are working on.
1. Are you starting a new ADK Kotlin project? If yes, begin from current official documentation and the current release feed rather than freezing the 1.0 launch snippet.
2. Are you on 1.0.x with resumable workflow agents or MCP tools? If yes, prioritize a controlled 1.1.0 regression because the release notes include fixes in those areas.
3. Do you need Android AppFunctions? If no, avoid adding the integration just because it is new. If yes, define the minimum function surface and confirmation policy before wiring it to the agent.
4. Are you using an Android extension labeled beta? If yes, preserve that maturity label in product and engineering decisions even though ADK core is GA.
5. Did the upgrade change agent behavior? If yes, compare event persistence, session restoration, tool rejection and provider/session IDs before attributing the regression to the model.
REPAIR FLOW
A visual path from symptom to verified production.
Observe→
Protect→
Isolate→
Repair→
Test→
Publish→
Verify live
BEFORE YOU PASTE CODE
Protect the working site first.
Pin or deliberately manage framework versions through the project's normal Gradle/version-catalog policy and preserve a tested rollback commit.
Keep secrets and production credentials out of agent prompts, local skill files, logs and example code.
Require explicit authorization for sensitive AppFunctions and tool calls, and test denial/rejection paths as well as success paths.
Verify Android, JVM and server-side behavior separately when the project uses multiplatform code; a successful common-module compile is not a substitute for target-specific tests.
STOP AND GET HELP WHEN
Do not turn a repair into a larger outage.
Do not call ML Kit GenAI generally available merely because ADK for Kotlin 1.0 is GA; Google's launch post labels that Android extension beta.
Do not enable Android AppFunctions broadly without checking which functions are exposed, what data they can access and whether state-changing actions need user confirmation.
Do not interpret an MCP transport recommendation as permission to weaken authentication, network policy or tool authorization.
Do not upgrade a production agent without testing session resumability, tool failures and the app-specific workflows that depend on persisted state.
Do not claim ADK for Kotlin 1.1.0 makes an agent production-safe by itself; application authorization, data handling, model behavior, observability and rollback remain separate engineering responsibilities.
HOW GENESIS HANDLES IT
Diagnose the exact failure before choosing a repair.
Genesis separates the visible symptom from the underlying technical cause. Run the supported diagnostic first, review the evidence, and then use a matching repair only when the failure is actually verified.
Platform scope
Google ADK for Kotlin 1.1.0 · Kotlin · Android · JVM · Kotlin Multiplatform · AppFunctions · MCP
Category
AI agents · Kotlin/Android · Framework upgrade
Last updated
2026-09-25
REPAIR PROFILE
Know the complexity before you edit.
Difficulty
Intermediate
Change risk
Medium
These labels describe implementation complexity and blast radius, not a guaranteed repair time.
AUTHORITATIVE SOURCES
Verify time-sensitive platform details at the source.
Google describes ADK for Kotlin 1.0 as generally available and production-ready at the core framework level. Optional integrations can have different maturity; for example, the 1.0 launch post labels the ML Kit GenAI Android extension beta.
What is the current ADK for Kotlin release?
The official google/adk-kotlin release page lists v1.1.0, released September 15, 2026, as the latest release in the source checked for this guide.
What changed in 1.1.0 for Android?
The official release notes add AppFunctions guidance, an AppFunctionsToolset and an Android AppFunctions conversion layer, alongside other framework fixes.
Does 1.1.0 change MCP guidance?
The 1.1.0 release notes include a fix that keeps an agent turn alive when an MCP server rejects a request and a documentation change recommending Streamable HTTP over SSE for MCP connections.
Should every 1.0.x app upgrade immediately?
Not blindly. Review the release notes, upgrade in a reversible branch, and test the workflows your application actually uses. The fixes are meaningful, but dependency changes still require app-specific verification.
Can AI coding tools read current ADK documentation?
Google's ADK documentation publishes llms.txt and llms-full.txt indexes and documents an ADK Docs MCP Server that supported coding tools can use for current documentation context.
Start with a free diagnostic. If Genesis verifies a problem and a compatible treatment exists, continue to the matching Code Store product or repair path. If you cannot find the exact code you need, request it at GenesisCodeDoctor.com rather than forcing a generic snippet into the wrong platform.
