Google's Agent Development Kit 2.14 moves core modules out of beta
ADK Python 2.14.0 graduates execution environments, A2A and evaluation from experimental to stable, adds a Microsoft Teams runner, and ships a breaking change to its skill-script tool.
Google shipped Agent Development Kit (ADK) Python 2.14.0 on 7 November. The release note calls out what’s graduating from experimental to stable: execution environments (LocalEnvironment, DaytonaEnvironment, E2BEnvironment and friends), the A2A agent-to-agent protocol, and the evaluation framework, alongside 26 auth, environment, app, CLI and cache modules in total.
It also adds a TeamsRunner that connects an ADK agent to Microsoft Teams chats and channels, and an OpenAIDecisionsClassifier for routing workflow and tool-gating questions through OpenAI’s Decisions API. Google gives a specific performance number: a 30-node workflow’s 200-turn run is 23% faster with 9% lower peak memory than before.
from google.adk.environments import DaytonaEnvironmentThere’s a breaking change worth checking before upgrading: RunSkillScriptTool no longer accepts a raw shell command string, it now runs from validated paths and arguments instead. Enum tool arguments also change shape, arriving as enum members rather than raw values.
Why it matters: ADK is one of the more widely used open frameworks for building multi-agent systems, and graduating core modules out of experimental status means teams can build on execution sandboxes and evaluation without expecting the API to shift under them next release.
The caveat: the breaking changes apply on upgrade, so anything calling RunSkillScriptTool with a raw command string needs updating first.