Skip to content
Go AI SDK

Go AI SDK

Build Bedrock and MCP-powered AI workflows in Go with tool loops and streaming.
Get Started View on GitHub

Why this project

  • Bedrock provider with multiple auth modes (profile, env, static creds, custom provider)
  • MCP client module with stdio, SSE, and HTTP transport helpers
  • Tool loop agent that can run with tools enabled or disabled
  • Streaming APIs for both direct model calls and agent flows
  • Independent modules so you can go get only what you need

Modular Install

Terminal window
go get github.com/iamanishx/go-ai/agent
go get github.com/iamanishx/go-ai/provider/bedrock
go get github.com/iamanishx/go-ai/mcp

Quick Example

import (
    "github.com/iamanishx/go-ai/agent"
    "github.com/iamanishx/go-ai/provider"
    "github.com/iamanishx/go-ai/provider/bedrock"
)


bedrockProvider := bedrock.Create(bedrock.BedrockProviderSettings{
    Region:  "us-east-1",
    Profile: "myprofile",
})


toolAgent := agent.CreateToolLoopAgent(agent.ToolLoopAgentSettings{
    Model: bedrockProvider.Chat("anthropic.claude-3-sonnet-20240229-v1:0"),
    Tools: []provider.Tool{
        {
            Name:        "get_weather",
            Description: "Get weather for a location",
            Parameters: map[string]interface{}{
                "type": "object",
                "properties": map[string]interface{}{
                    "location": map[string]interface{}{"type": "string"},
                },
                "required": []string{"location"},
            },
        },
    },
    ExecuteTools: true,
})


result, err := toolAgent.Generate(ctx, agent.AgentCallOptions{
    Prompt: "What's the weather in San Francisco?",
})


if err != nil {
    panic(err)
}


_ = result.Text