Eino ADK: Quickstart

Installation

Eino provides ADK from v0.5.0. Upgrade your project:

// stable >= eino@v0.5.0
go get github.com/cloudwego/eino@latest

Agent

What is Eino ADK

Eino ADK, inspired by Google‑ADK, is a Go framework for building Agent and Multi‑Agent applications. It standardizes context passing, event streaming, task transfer, interrupts/resume, and cross‑cutting features.

What is an Agent

An Agent represents an executable, intelligent task unit with a clear name and description so other agents can discover and transfer tasks to it. Typical use cases:

Query weather information
Book meetings
Answer domain‑specific questions

Agent in ADK

All ADK features build on the Agent abstraction:

type Agent interface {
    Name(ctx context.Context) string
    Description(ctx context.Context) string
    Run(ctx context.Context, input *AgentInput) *AsyncIterator[*AgentEvent]
}

ADK provides three base categories:

ChatModel Agent — the “thinking” part powered by LLMs; understand, reason, plan, respond, and call tools
Workflow Agents — coordination layer with preset logic (sequential/parallel/loop). Deterministic, predictable flows.
- Sequential — execute subagents in order
- Loop — repeat subagents until a condition
- Parallel — run subagents concurrently
Custom Agent — implement the interface for bespoke logic

Combine these to compose Multi‑Agent systems. Eino also offers built‑in best‑practice paradigms:

Supervisor — centralized coordinator controlling communications and delegation
Plan‑Execute — planner generates steps; executor carries them out; replanner decides finish or replan

Category	ChatModel Agent	Workflow Agents	Custom Logic	EinoBuiltInAgent (supervisor, plan‑execute)
Function	Thinking, generation, tool calls	Control execution flow among agents	Run custom logic	Out‑of‑the‑box multi‑agent patterns
Core	LLM	Predetermined flows (sequential/parallel/loop)	Custom code	High‑level encapsulation based on Eino practice
Purpose	Generation, dynamic decisions	Structured orchestration	Specific customization	Turnkey solutions for common scenarios

ADK Examples

Explore examples in Eino‑examples. The table summarizes project paths, key points, and diagrams:

Project Path	Intro	Diagram
Sequential workflow	This example shows a sequential multi‑agent workflow built with Eino ADK’s Workflow paradigm. Sequential construction: create a ResearchAgent via adk.NewSequentialAgent with two subagents — PlanAgent (planning) and WriterAgent (writing). Clear responsibilities: PlanAgent outputs a detailed plan; WriterAgent writes a structured report based on the plan. Chained IO: PlanAgent’s output feeds WriterAgent’s input, illustrating ordered dependency.
Loop workflow	Built with LoopAgent to form a reflection‑iteration framework. Iterative reflection: ReflectionAgent combines MainAgent (solve) and CritiqueAgent (review), up to 5 iterations. MainAgent: produces an initial solution. CritiqueAgent: audits quality, suggests improvements; terminates when satisfactory. Loop mechanism: repeatedly improves outputs across iterations.
Parallel workflow	Built with ParallelAgent for concurrent data collection. Concurrent framework: DataCollectionAgent launches multiple info collectors. Responsibility split: each subagent handles one channel independently. Parallel execution: starts tasks simultaneously to improve throughput.
supervisor	Single‑layer Supervisor manages two composite subagents: Research Agent (retrieval) and Math Agent (math operations: add/multiply/divide). All math ops are handled by one Math Agent rather than splitting into many; suitable for focused tasks and quick deployment.
layered‑supervisor	Multi‑tier supervision: top Supervisor manages Research Agent and Math Agent; Math Agent further manages Subtract/Multiply/Divide subagents. Top Supervisor delegates research/math tasks. Mid‑tier Math Agent delegates specific operations. Good for fine‑grained decomposition and multi‑level delegation.
plan‑execute example	Implements a plan‑execute‑replan travel planner: Planner generates stepwise plan; Executor calls mock tools (get_weather/search_flights/search_hotels/search_attractions/ask_for_clarification); Replanner decides replan or finish. Two layers: Layer 2: loop of execute + replan. Layer 1: sequential of plan + layer‑2 loop.
book recommendation agent (interrupt/resume)	Demonstrates a ChatModel agent with tools and checkpointing. Agent: BookRecommender via adk.NewChatModelAgent. Tools: BookSearch and AskForClarification. State: in‑memory checkpoint storage. Events: iterate runner.Query and runner.Resume. Custom input: drive flow via options.