ADK Go
A code-first Go toolkit for building, evaluating, orchestrating, and deploying sophisticated AI-agent workflows.
- Source repo
- google/adk-go
- Stars
- ★ 8.9k
- Last updated
- today
- License
- Apache-2.0
- Primary language
- Go
- FA score
- 31/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A cloud-native Go team that wants agent logic, tools, and orchestration maintained and versioned beside its existing Go services.
- Not a fit if
- Developers who want to build agents in a language other than Go
- Users wanting a ready-made agent with no code to write
- Source review
- 31/100 · Major gaps 7 safety controls not found
What does this agent do, and when should you use it?
ADK Go is Google's open-source Go toolkit for building, evaluating, and deploying sophisticated AI agents. Developers define agent logic, tools, and orchestration directly in Go, making that code testable and versionable. Agents can use pre-built tools, custom functions, or integrations with existing tools, while specialized agents can be composed into multi-agent systems. The project targets cloud-native applications, supports containerization, and calls out strong support for Google Cloud Run. It is optimized for Gemini, while the README describes it as model-agnostic, deployment-agnostic, and compatible with other frameworks.
After adding google.golang.org/adk/v2 to a Go project, developers can define agent logic, tools, and orchestration in Go. Tool capabilities may come from pre-built tools, custom functions, or integrations with existing tools, and specialized agents can be composed into a multi-agent application. The repository describes its scope as building, evaluating, deploying, and orchestrating agent workflows, including containerized deployment to cloud-native environments such as Google Cloud Run. The supplied material does not identify concrete Go types, functions, runtime entrypoints, I/O formats, or a runnable first example.
- A cloud-native Go team that wants agent logic, tools, and orchestration maintained and versioned beside its existing Go services.
- An application developer building a workflow by composing several specialized agents.
- An engineering team that needs to extend agents with pre-built tools, custom Go functions, or existing tool integrations.
- A team planning to containerize a Go agent application for deployment to Google Cloud Run or another cloud-native environment.
- A developer centered on Gemini who wants the stated flexibility to choose models and deployment targets.
How do you install or deploy this agent?
From a Go project directory, run:
go get google.golang.org/adk/v2The supplied material does not state the required Go version, credentials, environment variables, module prerequisites, or dependency setup details.
How do you use this agent?
The material says that agent logic, tools, and orchestration are defined in Go and that specialized agents can be composed, but it provides no copyable API example, first-run command, runtime entrypoint, credential configuration, or deployment configuration. A verified first invocation therefore cannot be derived from the supplied evidence alone.
What are this agent's strengths and limitations?
- Go-native development: agent logic, tools, and orchestration are defined directly in Go code.
- Explicit support for modular multi-agent systems made from specialized agents.
- Tool integration spans pre-built tools, custom functions, and existing tools.
- Designed for cloud-native deployment, with containerization and explicit strong support for Google Cloud Run.
- The supplied material provides no concrete API, runnable example, or runtime entrypoint, so initial integration requires additional documentation discovery.
- Required Go version, authentication, environment variables, and model connection settings are not stated.
- Although it is described as model-agnostic, no non-Gemini provider or adapter path is documented in the supplied evidence.
- Evaluation is named as a project goal, but no evaluation interface, metric, or execution flow is shown.
How does this agent compare with similar options?
The repository links Python, Java, Kotlin, and TypeScript versions of ADK, but the supplied material does not document functional or compatibility differences from the Go version.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| ADK Go This agent | 31 · Major gaps | Library / SDKFree + model costs | ★ 8.9k | today | Go | — |
| Google ADK for TypeScript | 51 · Major gaps | CLIFree + model costs | ★ 1.5k | today | TypeScript | — |
| AI Agent Engineering Textbook | 55 · Major gaps | FrameworkFree | ★ 651 | 13d ago | HTML | — |
| Hello-Agents | 54 · Major gaps | Web appFree + model costs | ★ 86k | 23d ago | Python | OpenAI API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows the repository is published by Google's official organization, but no specific implementations or documentation for permission management, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, or rollback mechanisms are provided. Therefore, except for source attribution, all other criteria are scored 0 due to insufficient evidence. Source attribution scores 2 because the publisher identity is clear and the repository ownership is unambiguous.
Self-consistency scores 2 because README and go.mod are consistent in version and module path. Dependency availability scores 2 because go.mod lists many dependencies, but no availability guarantees are provided. Failure messages scores 1 because README mentions error handling in mock contexts, but no comprehensive failure message strategy is provided.
Audience and scenarios scores 1 because README mentions an Agent development toolkit for Go developers, but does not detail the target audience. Capability boundaries scores 1 because README mentions breaking changes, but does not clearly define capability boundaries. Trigger precision scores 1 because README mentions the use of context.Context, but does not detail trigger conditions. Environment fit scores 1 because go.mod and CI configuration indicate Go environment support, but no other environments are detailed.
Information architecture scores 1 because README provides breaking changes information, but no complete documentation structure. Install notes scores 1 because go.mod provides module path, but no installation steps. Naming stability scores 1 because README mentions version 2.0, but no naming conventions. Examples and FAQ scores 1 because README provides code examples, but no FAQ. Known limitations scores 1 because README mentions breaking changes, but does not explicitly list known limitations. License scores 3 because LICENSE file fully provides Apache 2.0 license. Versioning and changelog scores 1 because README mentions version 2.0, but no changelog. Maintenance responsibility scores 2 because publisher identity is clear, but no maintenance plan is provided.
Output usability scores 1 because README provides code examples, but no complete output format. Marginal value scores 1 because the toolkit provides Agent development functionality, but no comparison with other tools. Cost-benefit scores 1 because it is open-source and free, but no performance or resource consumption information is provided.
Claim traceability scores 1 because README mentions PR links, but no complete claim sources. Cross-source corroboration scores 1 because go.mod and CI configuration corroborate each other, but no external verification. Fact-inference separation scores 1 because README distinguishes breaking changes and migration advice, but does not explicitly separate facts and inferences.
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Static review cannot verify actual runtime behavior; all conclusions about security, reliability, and effectiveness are inferred from file contents.
- The repository does not provide specific implementations or documentation for security features such as permission management, user confirmation, and data flow transparency; evaluate before use.
- Many dependencies are present without a security audit report; it is recommended to run tools like govulncheck to check for dependency vulnerabilities before use.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
plugin/agentanalytics/v0.1.0 plugin/agentanalytics v0.1.0
Release notes excerpt
The first release of the BigQuery Agent Analytics plugin for ADK Go. It logs every agent run to a BigQuery table, so you can query and dashboard agent behavior with SQL: invocations, model requests and responses, tool calls, state changes, agent transfers and human-in-the-loop requests. The table schema, event type names and analytics views match adk-python's BigQuery Agent Analytics plugin, so Go and Python agents can write to the same table and be queried together. Install ```bash go get google.golang.org/adk/plugin/[email protected] ``` …View on GitHub ↗ -
Release notes excerpt
[2.7.1](https://github.com/google/adk-go/compare/v2.7.0...v2.7.1) (2026-11-04) Bug Fixes * agentanalytics: classify Write API errors and split oversized batches ([#2016](https://github.com/google/adk-go/issues/2016)) ([2466e97](https://github.com/google/adk-go/commit/2466e976ac371b8e2074315c70cbbdaf9b1f2498)) * model/openaimodel: refuse media in system and developer Responses messages ([#2025](https://github.com/google/adk-go/issues/2025)) ([40f132d](https://github.com/google/adk-go/commit/40f132d697752596c79d307c4ef2bfb70c55ad9c)) …View on GitHub ↗ -
Release notes excerpt
Highlights v2.7.0 adds Anthropic Claude as a first-party model, search-gated tool discovery for large tool catalogs, an experimental classifier API for workflow routing, and a per-invocation budget on model calls. It also switches telemetry to the OpenTelemetry GenAI semantic conventions by default, which is a breaking change for anyone consuming ADK traces. …View on GitHub ↗
FAQ
Is Gemini the only model it can use?
Can it be deployed on Google Cloud Run?
What credentials are required?
What license applies?
internal/httprr as an exception with its own license file.