Connect to an MCP server
You connect one MCP client over the transport that fits your deployment. Streamable HTTP is the normal remote choice and supports both stateful and stateless servers.
Unit example (nearest native match)
See the idea in context
transport, err := ax.NewAxMCPStreamableHTTPTransport(endpoint, nil)
if err != nil { panic(err) }
mcp := ax.NewAxMCPClient(transport, map[string]ax.Value{"namespace": "inventory"})
defer func() { _ = mcp.Close() }()
catalog, err := mcp.InspectCatalog(false)
if err != nil { panic(err) }
fmt.Printf("MCP catalog: %d tools, %d resources, %d templates\n",
len(catalog.Tools), len(catalog.Resources), len(catalog.ResourceTemplates))
program := ax.NewAx("request:string -> answer:string", map[string]ax.Value{"mcp": mcp})
llm := ax.NewOpenAICompatibleClient(map[string]ax.Value{
"api_key": key, "model": "gpt-5.4-mini",
})
output, err := program.Forward(context.Background(), llm,
map[string]ax.Value{"request": "Reindex inventory."}, nil)
if err != nil { panic(err) }
fmt.Println(output)- Create the transport
Streamable HTTP connects the client to the remote URL.
- Create one client
AxMCPClient owns era classification and the protocol lifecycle.
- Use a namespace
orders keeps discovered names clear when several servers are attached.
Run itIn your own project
go get github.com/ax-llm/ax/packages/go
import axllm "github.com/ax-llm/ax/packages/go"
client := axllm.NewAI("openai", map[string]axllm.Value{"apiKey": os.Getenv("OPENAI_API_KEY")})
classify := axllm.NewAx("review:string -> sentiment:class \"positive, negative, neutral\"", nil)Set OPENAI_APIKEY in your environment before running provider-backed code.
From a clone of the ax repo:
npm run example -- go src/examples/go/mcp/native_mcp_tools.goActive practice
Answer 2 in a row to learn this · attempt 1