crush/internal/agent/tools/mcp/init.go

// Package mcp provides functionality for managing Model Context Protocol (MCP)
// clients within the Crush application.
package mcp

import (
	"cmp"
	"context"
	"errors"
	"fmt"
	"io"
	"log/slog"
	"maps"
	"net/http"
	"os"
	"os/exec"
	"strings"
	"sync"
	"time"

	"github.com/charmbracelet/crush/internal/config"
	"github.com/charmbracelet/crush/internal/csync"
	"github.com/charmbracelet/crush/internal/home"
	"github.com/charmbracelet/crush/internal/permission"
	"github.com/charmbracelet/crush/internal/pubsub"
	"github.com/charmbracelet/crush/internal/version"
	"github.com/modelcontextprotocol/go-sdk/mcp"
)

var (
	sessions = csync.NewMap[string, *mcp.ClientSession]()
	states   = csync.NewMap[string, ClientInfo]()
	broker   = pubsub.NewBroker[Event]()
)

// State represents the current state of an MCP client
type State int

const (
	StateDisabled State = iota
	StateStarting
	StateConnected
	StateError
)

func (s State) String() string {
	switch s {
	case StateDisabled:
		return "disabled"
	case StateStarting:
		return "starting"
	case StateConnected:
		return "connected"
	case StateError:
		return "error"
	default:
		return "unknown"
	}
}

// EventType represents the type of MCP event
type EventType uint

const (
	EventStateChanged EventType = iota
	EventToolsListChanged
	EventPromptsListChanged
)

// Event represents an event in the MCP system
type Event struct {
	Type   EventType
	Name   string
	State  State
	Error  error
	Counts Counts
}

// Counts number of available tools, prompts, etc.
type Counts struct {
	Tools   int
	Prompts int
}

// ClientInfo holds information about an MCP client's state
type ClientInfo struct {
	Name        string
	State       State
	Error       error
	Client      *mcp.ClientSession
	Counts      Counts
	ConnectedAt time.Time
}

// SubscribeEvents returns a channel for MCP events
func SubscribeEvents(ctx context.Context) <-chan pubsub.Event[Event] {
	return broker.Subscribe(ctx)
}

// GetStates returns the current state of all MCP clients
func GetStates() map[string]ClientInfo {
	return maps.Collect(states.Seq2())
}

// GetState returns the state of a specific MCP client
func GetState(name string) (ClientInfo, bool) {
	return states.Get(name)
}

// Close closes all MCP clients. This should be called during application shutdown.
func Close() error {
	var errs []error
	for name, c := range sessions.Seq2() {
		if err := c.Close(); err != nil &&
			!errors.Is(err, io.EOF) &&
			!errors.Is(err, context.Canceled) &&
			err.Error() != "signal: killed" {
			errs = append(errs, fmt.Errorf("close mcp: %s: %w", name, err))
		}
	}
	broker.Shutdown()
	return errors.Join(errs...)
}

// Initialize initializes MCP clients based on the provided configuration.
func Initialize(ctx context.Context, permissions permission.Service, cfg *config.Config) {
	var wg sync.WaitGroup
	// Initialize states for all configured MCPs
	for name, m := range cfg.MCP {
		if m.Disabled {
			updateState(name, StateDisabled, nil, nil, Counts{})
			slog.Debug("skipping disabled mcp", "name", name)
			continue
		}

		// Set initial starting state
		updateState(name, StateStarting, nil, nil, Counts{})

		wg.Add(1)
		go func(name string, m config.MCPConfig) {
			defer func() {
				wg.Done()
				if r := recover(); r != nil {
					var err error
					switch v := r.(type) {
					case error:
						err = v
					case string:
						err = fmt.Errorf("panic: %s", v)
					default:
						err = fmt.Errorf("panic: %v", v)
					}
					updateState(name, StateError, err, nil, Counts{})
					slog.Error("panic in mcp client initialization", "error", err, "name", name)
				}
			}()

			ctx, cancel := context.WithTimeout(ctx, mcpTimeout(m))
			defer cancel()

			session, err := createSession(ctx, name, m, cfg.Resolver())
			if err != nil {
				return
			}

			tools, err := getTools(ctx, session)
			if err != nil {
				slog.Error("error listing tools", "error", err)
				updateState(name, StateError, err, nil, Counts{})
				session.Close()
				return
			}

			prompts, err := getPrompts(ctx, session)
			if err != nil {
				slog.Error("error listing prompts", "error", err)
				updateState(name, StateError, err, nil, Counts{})
				session.Close()
				return
			}

			updateTools(name, tools)
			updatePrompts(name, prompts)
			sessions.Set(name, session)

			updateState(name, StateConnected, nil, session, Counts{
				Tools:   len(tools),
				Prompts: len(prompts),
			})
		}(name, m)
	}
	wg.Wait()
}

func getOrRenewClient(ctx context.Context, name string) (*mcp.ClientSession, error) {
	sess, ok := sessions.Get(name)
	if !ok {
		return nil, fmt.Errorf("mcp '%s' not available", name)
	}

	cfg := config.Get()
	m := cfg.MCP[name]
	state, _ := states.Get(name)

	timeout := mcpTimeout(m)
	pingCtx, cancel := context.WithTimeout(ctx, timeout)
	defer cancel()
	err := sess.Ping(pingCtx, nil)
	if err == nil {
		return sess, nil
	}
	updateState(name, StateError, maybeTimeoutErr(err, timeout), nil, state.Counts)

	sess, err = createSession(ctx, name, m, cfg.Resolver())
	if err != nil {
		return nil, err
	}

	updateState(name, StateConnected, nil, sess, state.Counts)
	sessions.Set(name, sess)
	return sess, nil
}

// updateState updates the state of an MCP client and publishes an event
func updateState(name string, state State, err error, client *mcp.ClientSession, counts Counts) {
	info := ClientInfo{
		Name:   name,
		State:  state,
		Error:  err,
		Client: client,
		Counts: counts,
	}
	switch state {
	case StateConnected:
		info.ConnectedAt = time.Now()
	case StateError:
		updateTools(name, nil)
		sessions.Del(name)
	}
	states.Set(name, info)

	// Publish state change event
	broker.Publish(pubsub.UpdatedEvent, Event{
		Type:   EventStateChanged,
		Name:   name,
		State:  state,
		Error:  err,
		Counts: counts,
	})
}

func createSession(ctx context.Context, name string, m config.MCPConfig, resolver config.VariableResolver) (*mcp.ClientSession, error) {
	timeout := mcpTimeout(m)
	mcpCtx, cancel := context.WithCancel(ctx)
	cancelTimer := time.AfterFunc(timeout, cancel)

	transport, err := createTransport(mcpCtx, m, resolver)
	if err != nil {
		updateState(name, StateError, err, nil, Counts{})
		slog.Error("error creating mcp client", "error", err, "name", name)
		cancel()
		cancelTimer.Stop()
		return nil, err
	}

	client := mcp.NewClient(
		&mcp.Implementation{
			Name:    "crush",
			Version: version.Version,
			Title:   "Crush",
		},
		&mcp.ClientOptions{
			ToolListChangedHandler: func(context.Context, *mcp.ToolListChangedRequest) {
				broker.Publish(pubsub.UpdatedEvent, Event{
					Type: EventToolsListChanged,
					Name: name,
				})
			},
			PromptListChangedHandler: func(context.Context, *mcp.PromptListChangedRequest) {
				broker.Publish(pubsub.UpdatedEvent, Event{
					Type: EventPromptsListChanged,
					Name: name,
				})
			},
			LoggingMessageHandler: func(_ context.Context, req *mcp.LoggingMessageRequest) {
				slog.Info("mcp log", "name", name, "data", req.Params.Data)
			},
			KeepAlive: time.Minute * 10,
		},
	)

	session, err := client.Connect(mcpCtx, transport, nil)
	if err != nil {
		err = maybeStdioErr(err, transport)
		updateState(name, StateError, maybeTimeoutErr(err, timeout), nil, Counts{})
		slog.Error("error starting mcp client", "error", err, "name", name)
		cancel()
		cancelTimer.Stop()
		return nil, err
	}

	cancelTimer.Stop()
	slog.Info("Initialized mcp client", "name", name)
	return session, nil
}

// maybeStdioErr if a stdio mcp prints an error in non-json format, it'll fail
// to parse, and the cli will then close it, causing the EOF error.
// so, if we got an EOF err, and the transport is STDIO, we try to exec it
// again with a timeout and collect the output so we can add details to the
// error.
// this happens particularly when starting things with npx, e.g. if node can't
// be found or some other error like that.
func maybeStdioErr(err error, transport mcp.Transport) error {
	if !errors.Is(err, io.EOF) {
		return err
	}
	ct, ok := transport.(*mcp.CommandTransport)
	if !ok {
		return err
	}
	if err2 := stdioCheck(ct.Command); err2 != nil {
		err = errors.Join(err, err2)
	}
	return err
}

func maybeTimeoutErr(err error, timeout time.Duration) error {
	if errors.Is(err, context.Canceled) {
		return fmt.Errorf("timed out after %s", timeout)
	}
	return err
}

func createTransport(ctx context.Context, m config.MCPConfig, resolver config.VariableResolver) (mcp.Transport, error) {
	switch m.Type {
	case config.MCPStdio:
		command, err := resolver.ResolveValue(m.Command)
		if err != nil {
			return nil, fmt.Errorf("invalid mcp command: %w", err)
		}
		if strings.TrimSpace(command) == "" {
			return nil, fmt.Errorf("mcp stdio config requires a non-empty 'command' field")
		}
		cmd := exec.CommandContext(ctx, home.Long(command), m.Args...)
		cmd.Env = append(os.Environ(), m.ResolvedEnv()...)
		return &mcp.CommandTransport{
			Command: cmd,
		}, nil
	case config.MCPHttp:
		if strings.TrimSpace(m.URL) == "" {
			return nil, fmt.Errorf("mcp http config requires a non-empty 'url' field")
		}
		client := &http.Client{
			Transport: &headerRoundTripper{
				headers: m.ResolvedHeaders(),
			},
		}
		return &mcp.StreamableClientTransport{
			Endpoint:   m.URL,
			HTTPClient: client,
		}, nil
	case config.MCPSSE:
		if strings.TrimSpace(m.URL) == "" {
			return nil, fmt.Errorf("mcp sse config requires a non-empty 'url' field")
		}
		client := &http.Client{
			Transport: &headerRoundTripper{
				headers: m.ResolvedHeaders(),
			},
		}
		return &mcp.SSEClientTransport{
			Endpoint:   m.URL,
			HTTPClient: client,
		}, nil
	default:
		return nil, fmt.Errorf("unsupported mcp type: %s", m.Type)
	}
}

type headerRoundTripper struct {
	headers map[string]string
}

func (rt headerRoundTripper) RoundTrip(req *http.Request) (*http.Response, error) {
	for k, v := range rt.headers {
		req.Header.Set(k, v)
	}
	return http.DefaultTransport.RoundTrip(req)
}

func mcpTimeout(m config.MCPConfig) time.Duration {
	return time.Duration(cmp.Or(m.Timeout, 15)) * time.Second
}

func stdioCheck(old *exec.Cmd) error {
	ctx, cancel := context.WithTimeout(context.Background(), time.Second*5)
	defer cancel()
	cmd := exec.CommandContext(ctx, old.Path, old.Args...)
	cmd.Env = old.Env
	out, err := cmd.CombinedOutput()
	if err == nil || errors.Is(ctx.Err(), context.DeadlineExceeded) {
		return nil
	}
	return fmt.Errorf("%w: %s", err, string(out))
}