Getting Started

Languages: English (current) · Português

This guide takes you from zero to a working crew.

Prerequisites

• Go 1.24+ — check with go version.
• An API key for an LLM provider (e.g. OPENAI_API_KEY) — or use an offline/custom LLM.

1. Create a project

mkdir my-crew && cd my-crew
go mod init example.com/my-crew
go get github.com/rhgs/crewai-go@latest

2. Write the program

main.go:

package main

import (
	"context"
	"fmt"
	"log"

	"github.com/rhgs/crewai-go"
	"github.com/rhgs/crewai-go/llm/openai"
)

func main() {
	llm := openai.New("gpt-4o-mini")

	researcher := crewai.NewAgent(
		"Researcher",
		"Find relevant, reliable information",
		"You are an experienced, skeptical analyst.",
		llm,
	)

	task := crewai.NewTask(
		"Explain in 3 points why Go is good for back-end.",
		"A list of 3 short items.",
		researcher,
	)

	crew := crewai.NewCrew([]*crewai.Agent{researcher}, []*crewai.Task{task})
	crew.Verbose = true // or use WithLogger for structured logging (see below)

	out, err := crew.Kickoff(context.Background(), nil)
	if err != nil {
		log.Fatal(err)
	}
	fmt.Println(out.Final)
}

Tip: For structured logging via log/slog, replace crew.Verbose = true with crew.WithLogger(slog.New(slog.NewJSONHandler(os.Stdout, &slog.HandlerOptions{Level: slog.LevelDebug}))). See Logging for details.

3. Run it

export OPENAI_API_KEY=sk-...
go run .

4. No API key? Run offline

You can implement the crewai.LLM interface yourself (see llms.md) or use the mock LLM from the test package. The examples/custom_llm example runs fully offline:

go run github.com/rhgs/crewai-go/examples/custom_llm

Next steps

• Agents — configure roles, goals, tools, and the agentic loop.
• Tasks — chain tasks with context, Async waves, structured output, and warnings.
• Crews — sequential, hierarchical, staged, and async-wave scheduling; progress callbacks.
• Memory — short-term bag, MemoryStore, FileStore, embeddings.
• Tools — give "hands" to your agents (including web search).
• LLMs — providers, native tool calling, and logging.
• MCP — connect external Model Context Protocol servers.