Attach mcp stdio box to agent

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Signed-off-by: mudler <mudler@localai.io>

core/agent/agent.go

@@ -241,6 +241,7 @@ func (a *Agent) Stop() {
 	a.Lock()
 	defer a.Unlock()
 	xlog.Debug("Stopping agent", "agent", a.Character.Name)
+	a.closeMCPSTDIOServers()
 	a.context.Cancel()
 }
 

core/agent/mcp.go

@@ -3,12 +3,14 @@ package agent
 import (
 	"context"
 	"encoding/json"
-	"errors"
 
 	mcp "github.com/metoro-io/mcp-golang"
 	"github.com/metoro-io/mcp-golang/transport/http"
+	stdioTransport "github.com/metoro-io/mcp-golang/transport/stdio"
 	"github.com/mudler/LocalAGI/core/types"
+	"github.com/mudler/LocalAGI/pkg/stdio"
 	"github.com/mudler/LocalAGI/pkg/xlog"
+
 	"github.com/sashabaranov/go-openai/jsonschema"
 )
 
@@ -19,6 +21,12 @@ type MCPServer struct {
 	Token string `json:"token"`
 }
 
+type MCPSTDIOServer struct {
+	Args []string `json:"args"`
+	Env  []string `json:"env"`
+	Cmd  string   `json:"cmd"`
+}
+
 type mcpAction struct {
 	mcpClient       *mcp.Client
 	inputSchema     ToolInputSchema
@@ -79,6 +87,68 @@ type ToolInputSchema struct {
 	Required   []string               `json:"required,omitempty"`
 }
 
+func (a *Agent) addTools(client *mcp.Client) (types.Actions, error) {
+
+	var generatedActions types.Actions
+	xlog.Debug("Initializing client")
+	// Initialize the client
+	response, e := client.Initialize(a.context)
+	if e != nil {
+		xlog.Error("Failed to initialize client", "error", e.Error())
+		return nil, e
+	}
+
+	xlog.Debug("Client initialized: %v", response.Instructions)
+
+	var cursor *string
+	for {
+		tools, err := client.ListTools(a.context, cursor)
+		if err != nil {
+			xlog.Error("Failed to list tools", "error", err.Error())
+			return nil, err
+		}
+
+		for _, t := range tools.Tools {
+			desc := ""
+			if t.Description != nil {
+				desc = *t.Description
+			}
+
+			xlog.Debug("Tool", "name", t.Name, "description", desc)
+
+			dat, err := json.Marshal(t.InputSchema)
+			if err != nil {
+				xlog.Error("Failed to marshal input schema", "error", err.Error())
+			}
+
+			xlog.Debug("Input schema", "tool", t.Name, "schema", string(dat))
+
+			// XXX: This is a wild guess, to verify (data types might be incompatible)
+			var inputSchema ToolInputSchema
+			err = json.Unmarshal(dat, &inputSchema)
+			if err != nil {
+				xlog.Error("Failed to unmarshal input schema", "error", err.Error())
+			}
+
+			// Create a new action with Client + tool
+			generatedActions = append(generatedActions, &mcpAction{
+				mcpClient:       client,
+				toolName:        t.Name,
+				inputSchema:     inputSchema,
+				toolDescription: desc,
+			})
+		}
+
+		if tools.NextCursor == nil {
+			break // No more pages
+		}
+		cursor = tools.NextCursor
+	}
+
+	return generatedActions, nil
+
+}
+
 func (a *Agent) initMCPActions() error {
 
 	a.mcpActions = nil
@@ -86,6 +156,7 @@ func (a *Agent) initMCPActions() error {
 
 	generatedActions := types.Actions{}
 
+	// MCP HTTP Servers
 	for _, mcpServer := range a.options.mcpServers {
 		transport := http.NewHTTPClientTransport("/mcp")
 		transport.WithBaseURL(mcpServer.URL)
@@ -95,70 +166,50 @@ func (a *Agent) initMCPActions() error {
 
 		// Create a new client
 		client := mcp.NewClient(transport)
+		xlog.Debug("Adding tools for MCP server", "server", mcpServer)
+		generatedActions, err = a.addTools(client)
+		if err != nil {
+			xlog.Error("Failed to add tools for MCP server", "server", mcpServer, "error", err.Error())
+		}
+	}
 
-		xlog.Debug("Initializing client", "server", mcpServer.URL)
+	// MCP STDIO Servers
-		// Initialize the client
+	a.closeMCPSTDIOServers() // Make sure we stop all previous servers if any is active
-		response, e := client.Initialize(a.context)
+	for _, mcpStdioServer := range a.options.mcpStdioServers {
-		if e != nil {
+		client := stdio.NewClient(a.options.mcpBoxURL)
-			xlog.Error("Failed to initialize client", "error", e.Error(), "server", mcpServer)
+		p, err := client.CreateProcess(a.context,
-			if err == nil {
+			mcpStdioServer.Cmd,
-				err = e
+			mcpStdioServer.Args,
-			} else {
+			mcpStdioServer.Env,
-				err = errors.Join(err, e)
+			a.Character.Name)
-			}
+		if err != nil {
+			xlog.Error("Failed to create process", "error", err.Error())
+			continue
+		}
+		read, writer, err := client.GetProcessIO(p.ID)
+		if err != nil {
+			xlog.Error("Failed to get process IO", "error", err.Error())
 			continue
 		}
 
-		xlog.Debug("Client initialized: %v", response.Instructions)
+		transport := stdioTransport.NewStdioServerTransportWithIO(read, writer)
 
-		var cursor *string
+		// Create a new client
-		for {
+		mcpClient := mcp.NewClient(transport)
-			tools, err := client.ListTools(a.context, cursor)
-			if err != nil {
-				xlog.Error("Failed to list tools", "error", err.Error())
-				return err
-			}
 
-			for _, t := range tools.Tools {
+		xlog.Debug("Adding tools for MCP server (stdio)", "server", mcpStdioServer)
-				desc := ""
+		generatedActions, err = a.addTools(mcpClient)
-				if t.Description != nil {
+		if err != nil {
-					desc = *t.Description
+			xlog.Error("Failed to add tools for MCP server", "server", mcpStdioServer, "error", err.Error())
-				}
-
-				xlog.Debug("Tool", "mcpServer", mcpServer, "name", t.Name, "description", desc)
-
-				dat, err := json.Marshal(t.InputSchema)
-				if err != nil {
-					xlog.Error("Failed to marshal input schema", "error", err.Error())
-				}
-
-				xlog.Debug("Input schema", "mcpServer", mcpServer, "tool", t.Name, "schema", string(dat))
-
-				// XXX: This is a wild guess, to verify (data types might be incompatible)
-				var inputSchema ToolInputSchema
-				err = json.Unmarshal(dat, &inputSchema)
-				if err != nil {
-					xlog.Error("Failed to unmarshal input schema", "error", err.Error())
-				}
-
-				// Create a new action with Client + tool
-				generatedActions = append(generatedActions, &mcpAction{
-					mcpClient:       client,
-					toolName:        t.Name,
-					inputSchema:     inputSchema,
-					toolDescription: desc,
-				})
-			}
-
-			if tools.NextCursor == nil {
-				break // No more pages
-			}
-			cursor = tools.NextCursor
 		}
-
 	}
 
 	a.mcpActions = generatedActions
 
 	return err
 }
+
+func (a *Agent) closeMCPSTDIOServers() {
+	client := stdio.NewClient(a.options.mcpBoxURL)
+	client.StopGroup(a.Character.Name)
+}

core/agent/options.go

@@ -50,7 +50,9 @@ type options struct {
 
 	conversationsPath string
 
-	mcpServers []MCPServer
+	mcpServers      []MCPServer
+	mcpStdioServers []MCPSTDIOServer
+	mcpBoxURL       string
 
 	newConversationsSubscribers []func(openai.ChatCompletionMessage)
 
@@ -207,6 +209,20 @@ func WithMCPServers(servers ...MCPServer) Option {
 	}
 }
 
+func WithMCPSTDIOServers(servers ...MCPSTDIOServer) Option {
+	return func(o *options) error {
+		o.mcpStdioServers = servers
+		return nil
+	}
+}
+
+func WithMCPBoxURL(url string) Option {
+	return func(o *options) error {
+		o.mcpBoxURL = url
+		return nil
+	}
+}
+
 func WithLLMAPIURL(url string) Option {
 	return func(o *options) error {
 		o.LLMAPI.APIURL = url

main.go

@@ -23,6 +23,7 @@ var apiKeysEnv = os.Getenv("LOCALAGI_API_KEYS")
 var imageModel = os.Getenv("LOCALAGI_IMAGE_MODEL")
 var conversationDuration = os.Getenv("LOCALAGI_CONVERSATION_DURATION")
 var localOperatorBaseURL = os.Getenv("LOCALOPERATOR_BASE_URL")
+var mcpboxURL = os.Getenv("LOCALAGI_MCPBOX_URL")
 
 func init() {
 	if baseModel == "" {