Back to Gap Map Product / UX / Orchestration & agents

Google Agent Development Kit (ADK)

Google
open source / Overall score: 4.3(strong)

Google's Agent Development Kit (ADK) is a code-first Python framework for building, evaluating and deploying AI agents. Agents carry instructions and tools, compose into multi-agent hierarchies, and run inside a graph-based workflow runtime with routing, loops, retries, shared state and human-in-the-loop steps. Tools can be Python functions, OpenAPI specs or MCP servers. Tuned for Gemini but model-agnostic, it has sibling versions in Java, Kotlin, Go and TypeScript.

Openness

5 high confidence
5.0
license
Apache-2.0 (OSI)
source
public
core features withheld
no — the tree at 9620b42 holds 3,084 paths under one root Apache-2.0 LICENSE with no ee/, enterprise/, commercial/ or proprietary/ path; the README calls the framework open source and deployment-agnostic, and names Cloud Run and Google's hosted Agent Runtime as places to deploy to rather than parts of it
commercial
Gemini-Enterprise-Agent-Platform

ADK is Apache-2.0 licensed, and the whole framework builds from its public repository with no enterprise directory or paid edition. Google sells hosted places to run agents, but deploying there is optional, and the README names Cloud Run and other targets as alternatives.

Adoption

5 medium confidence
5.0

Adoption is read from PyPI downloads of google-adk over the trailing month. CI pipelines and dependent packages add to the count, so it measures installs rather than developers, and the Java, Kotlin, Go and TypeScript versions are not included.

Capability

4 medium confidence
4.0

ADK pairs multi-agent hierarchies with a graph workflow runtime that handles routing, loops, retries and human approval. It sits a step below LangGraph, which documents durable runs that resume exactly where they stopped; the ADK README makes no such claim.

  • https://raw.githubusercontent.com/google/adk-python/main/README.md recorded 2026-11-12

    The README lists a "Workflow Runtime" with "routing, fan-out/fan-in, loops, retry, state management, dynamic nodes, human-in-the-loop, and nested workflows", a Task API for agent-to-agent delegation, multi-agent hierarchies, MCP and OpenAPI tools and a tool confirmation flow; it publishes no benchmark figure.

Verified 2026-11-12