Build streaming AI backends in Go with React compatibility

Grafana AI SDK gives Go applications one API for model calls, streaming, tools, structured output, and multi-step agents across supported providers. It follows the design of Vercel's AI SDK and stays wire-compatible with its TypeScript frontend hooks, so a Go endpoint can stream Server-Sent Events directly to hooks such as useChat without a protocol adapter. StreamText and GenerateText stream a response or wait for the complete result, with retries and multi-step tool execution; the SDK serves useChat, useCompletion, and useObject for React compatibility; tools are composable plain Go functions a model can call, with a required-approval path for consequential actions; and structured output generates schema-validated objects, arrays, and choices.

Use it to build an AI-powered backend in Go that either pairs with an existing AI SDK React frontend or replaces a TypeScript backend outright, when you want streaming, tool-calling, and structured output without hand-rolling the SSE protocol that useChat expects. It fits production services that need timeouts, fallback, logging, Prometheus metrics, and Agent Observability middleware built in. It is specific to Go backends speaking the AI SDK wire protocol, so it is not the right choice if your frontend or backend stack has no AI SDK React compatibility requirement and a provider's own native SDK would be simpler.

mkdir ai-sdk-quickstart
cd ai-sdk-quickstart
go mod init example.com/ai-sdk-quickstart
go get github.com/grafana/ai-sdk
go get github.com/grafana/ai-sdk/providers/anthropic

package main

import (
	"context"
	"fmt"
	"log"
	"os"

	aisdk "github.com/grafana/ai-sdk"
	"github.com/grafana/ai-sdk/provider"
	"github.com/grafana/ai-sdk/providers/anthropic"
)

func main() {
	apiKey := os.Getenv("ANTHROPIC_API_KEY")
	if apiKey == "" {
		log.Fatal("ANTHROPIC_API_KEY is required")
	}

	model := anthropic.New(apiKey, "claude-sonnet-5")
	result, err := aisdk.GenerateText(context.Background(), model,
		aisdk.WithModelMessages(provider.UserText("Explain goroutines in one sentence.")),
	)
	if err != nil {
		log.Fatal(err)
	}

	fmt.Println(result.Text)
}

Run it with ANTHROPIC_API_KEY=sk-... go run .. Input is a sequence of model messages built with helpers like provider.UserText; output is generated text, a stream of SSE events for a React client, or a schema-validated structured object depending on which function is called.

Supported providers include Anthropic, Amazon Bedrock, OpenAI, OpenAI-compatible APIs, and Grafana's own hosted endpoint for internal services. It integrates with an AI SDK React frontend via the same wire protocol used by useChat, useCompletion, and useObject, and ships production middleware for timeouts, fallback, logging, and Prometheus metrics, plus an Agent Observability middleware. Licensed under Apache License 2.0, following the design of Vercel's AI SDK, itself also Apache-2.0 licensed.

Go backend developers who want to call LLMs, stream responses, and expose tool-calling or structured-output endpoints that a React frontend can consume with the standard AI SDK hooks, without writing their own streaming protocol layer.