adk-go
Build and orchestrate AI agent workflows with a code-first Go toolkit.
Code-first agent development framework optimized for concurrency and high performance.
ADK Go is an open-source framework designed to build, evaluate, and deploy AI agents using Go. It suits cloud-native engineering teams building multi-agent systems who prefer writing explicit code over declarative YAML configs.
What it is
The framework provides a modular set of components to define agent logic, integrate pre-built tools, and orchestrate multi-agent workflows directly in Go. It includes machine-readable documentation indices for AI coding agents to assist in generating application logic.
Who it's for
It suits backend engineers and architects building cloud-native agent applications in Go, particularly those deploying to Google Cloud Run. It does not suit teams without Go expertise or those looking for visual, no-code agent builders.
Key features
Idiomatic Go Design
Build agents using native Go syntax and concurrency patterns.
Code-First Workflows
Define agent logic and orchestration directly in code for easier testing and version control.
Modular Multi-Agent Systems
Compose multiple specialized agents together to scale complex applications.
AI Coding Agent Support
Reference dedicated machine-readable documentation endpoints to guide LLMs in writing agent code.
Up and running in 3 steps
Our honest verdict
What's great
- Leverages Go's performance and concurrency strengths for cloud-native applications.
- Code-first approach allows traditional software testing and versioning.
- Optimized for Gemini while remaining model-agnostic and deployment-agnostic.
- Machine-readable documentation formats make it easy for AI coding assistants to generate valid implementations.
Where it falls short
- Requires writing code in Go 1.26.10 or later.
- Requires managing your own containerization and deployment pipelines, such as Google Cloud Run.
- Lacks visual or no-code interfaces for non-programmers.
Pricing & license
The toolkit is free under its open-source license. The only cost is running your own infrastructure and any underlying LLM API usage.
Open remote AI-training roles
880 open remote AI-training roles, updated November 15, 2026; most pay $20–$175 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
Use it, or have us set it up
FAQ
What Go version does ADK Go require?
It requires Go 1.26.10 or later.
Can I use models other than Gemini?
Yes, while optimized for Gemini, the framework is model-agnostic.
