Skip to content
Agent Radar

adk-python 2.12.0 adds tool output callbacks and approvals

Updated adk-python

Summary

Google's adk-python v2.12.0, released October 13, 2026, adds after_tool_callback, require_confirmation, custom Agent Engine service accounts and Python 3.15.

Tool
adk-python
Category
Frameworks
Date
Open source
Yes
License
Apache-2.0
Pricing
Not stated in the sources we cited.
Sources

Google published adk-python v2.12.0 on October 13, 2026. According to the release notes, the update speeds up long sessions and adds finer control over tool execution, workflow approvals and deployment security.

Release facts

FactDetail
Productadk-python (Agent Development Kit for Python)
Versionv2.12.0
Release dateOctober 13, 2026
LicenseApache-2.0, as reported by the GitHub API
Where to get itGitHub release v2.12.0

What changed

  • Performance: the release notes say time to first model request is cut by 40% in long sessions.
  • Tool output filtering: after_tool_callback lets you intercept and filter tool outputs before they return to the model.
  • Workflow approvals: require_confirmation requires human confirmation before workflow nodes execute, and is supported in @node and FunctionNode.
  • Deployment: agents can be deployed to Agent Engine using dedicated custom service accounts.
  • Python 3.15: the notes say ADK can be installed and run on Python 3.15.
  • Other additions: the feature list includes an --state_file option for adk run, encrypted Google credentials in sessions, and skill-script outputs persisted as artifacts.

Questions

Can tool output be screened before the model sees it? Yes. The release notes describe after_tool_callback for intercepting and filtering tool outputs before they go back to the model.

Can a workflow pause for human approval? According to the release notes, require_confirmation requires human confirmation before executing workflow nodes.

Is adk-python open source? The repository description calls it an open-source, code-first Python toolkit, and the GitHub API reports the Apache-2.0 license.

See the adk-python hub for more, browse other agent frameworks, compare with Mastra or OpenAI Agents, or revisit this adk-python v2.12.0 launch entry.

Alternatives

adk-python compared with its alternatives
ToolLicensePricing note
adk-pythonThis launchAn open-source, code-first Python toolkit for building, evaluating and deploying AI agents.
Apache-2.0Not stated
OpenAI AgentsOpenAI's agent harness, shipped as a Python SDK update in April 2026 and as the hosted Agents API in September 2026.
Open sourceSDK and Agents API use standard token and tool pricing. The Agents API has no additional fee.
MastraTypeScript agent framework. Mastra Code is the terminal agent; AgentController is the extracted loop, in @mastra/core 1.47.0 or later.
Open sourceMastra Code and AgentController are announced as open source. No license fee is stated in the posts we cited.
AgnoAgno is an orchestration and service layer for agents, exposing them through a REST API and an MCP server.
Apache-2.0Not stated
OpenHandsMIT-licensed project to build and run AI coding agents: start locally, automate workflows, use any model.
MITThe OpenHands repository is MIT-licensed. Hosted plans are not priced on this page.

Questions

What is adk-python?

The repository describes it as an open-source, code-first Python toolkit for building, evaluating, and deploying sophisticated AI agents.

Is adk-python open source?

Yes. The GitHub API reports the Apache-2.0 license for the repository.

What languages does ADK support?

The ADK site says it is available in Python, TypeScript, Go, Java, and Kotlin.

Filed on Agent Radar. Claims in this note are limited to the sources above.