feat: MCP-Server für remote_execute + Inline-Choice-UI

- Python MCP-Server (stdio, JSON-RPC 2.0) für remote_execute via Claude CLI
- Interner API-Endpoint /api/internal/remote-execute mit Token-Auth
- MCP-Config wird automatisch generiert wenn Remote-Verbindung vorhanden
- Auth-Middleware: /api/internal/ von Auth ausgenommen
- Frontend: :::ASK_USER_CHOICE::: Blöcke als interaktive Buttons rendern
- Streaming: Unvollständige Choice-Blöcke werden während des Streams versteckt
- tool-ui.js: MCP-Tool-Namen generisch normalisieren (mcp__server__tool → tool)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Endpoints/ChatEndpoint.cs

@@ -245,7 +245,7 @@ public static class ChatEndpoint
                 {
                     if (anthropic.IsCliAvailable)
                     {
-                        await anthropic.StreamChatViaCli(res, ctx.ModelId, ctx.SystemPrompt, messages, ctx.ChainId, workspace: ctx.Workspace, allowUserChoice: ctx.Agent?.AllowUserChoice ?? true, planMode: body.PlanMode, webSearch: !string.IsNullOrEmpty(ctx.WebSearchVersion), sessionId: sessionId, workCt: workCt, sseCt: sseCt);
+                        await anthropic.StreamChatViaCli(res, ctx.ModelId, ctx.SystemPrompt, messages, ctx.ChainId, workspace: ctx.Workspace, allowUserChoice: ctx.Agent?.AllowUserChoice ?? true, planMode: body.PlanMode, webSearch: !string.IsNullOrEmpty(ctx.WebSearchVersion), sessionId: sessionId, remoteMgr: remoteMgr, workCt: workCt, sseCt: sseCt);
                     }
                     else
                     {

Endpoints/InternalEndpoint.cs

@@ -0,0 +1,73 @@
+using System.IO;
+using System.Linq;
+using System.Text.Json;
+using System.Threading;
+using System.Threading.Tasks;
+using Gott.Services;
+using Microsoft.AspNetCore.Builder;
+using Microsoft.AspNetCore.Http;
+using Microsoft.AspNetCore.Routing;
+
+namespace Gott.Endpoints;
+
+/// <summary>
+/// Interne API-Endpoints für MCP-Server und andere lokale Integrationen.
+/// Gesichert über einen internen Token (da Reverse-Proxy den Localhost-Check umgeht).
+/// </summary>
+public static class InternalEndpoint
+{
+    /// <summary>
+    /// Interner Auth-Token – wird beim Start generiert und an MCP-Server übergeben.
+    /// </summary>
+    public static string InternalToken { get; } = System.Guid.NewGuid().ToString("N");
+
+    public static void MapInternalEndpoints(this IEndpointRouteBuilder app)
+    {
+        app.MapPost("/api/internal/remote-execute", HandleRemoteExecute);
+    }
+
+    private static async Task<IResult> HandleRemoteExecute(HttpContext ctx, RemoteConnectionManager mgr)
+    {
+        try
+        {
+            // Token-Check: Header "X-Internal-Token" muss den internen Token enthalten
+            var token = ctx.Request.Headers["X-Internal-Token"].FirstOrDefault() ?? "";
+            if (token != InternalToken)
+                return Results.StatusCode(403);
+
+            using var reader = new StreamReader(ctx.Request.Body);
+            var body = await reader.ReadToEndAsync();
+
+            JsonElement json;
+            try { json = JsonSerializer.Deserialize<JsonElement>(body); }
+            catch { return Results.BadRequest(new { error = "Invalid JSON" }); }
+
+            var command = json.TryGetProperty("command", out var cmd) ? cmd.GetString() ?? "" : "";
+            var sessionId = json.TryGetProperty("sessionId", out var sid) ? sid.GetString() ?? "" : "";
+
+            if (string.IsNullOrEmpty(command))
+                return Results.BadRequest(new { error = "command erforderlich" });
+
+            if (string.IsNullOrEmpty(sessionId))
+                return Results.BadRequest(new { error = "sessionId erforderlich" });
+
+            var conn = mgr.GetBySessionId(sessionId);
+            if (conn == null)
+                return Results.NotFound(new { error = "Keine Remote-Verbindung für diese Session" });
+
+            using var cts = new CancellationTokenSource(System.TimeSpan.FromSeconds(180));
+            var result = await mgr.ExecuteCommand(conn.Id, command, 180, cts.Token);
+
+            return Results.Ok(new
+            {
+                exitCode = result.ExitCode,
+                stdout = result.Output ?? "",
+                stderr = result.Error ?? ""
+            });
+        }
+        catch (System.Exception ex)
+        {
+            return Results.Json(new { error = ex.Message, stackTrace = ex.StackTrace }, statusCode: 500);
+        }
+    }
+}

Middleware/AuthMiddleware.cs

@@ -26,7 +26,8 @@ public class AuthMiddleware
         var isImageFile = path.StartsWith("/api/images/file/") && ctx.Request.Method == "GET";
         var isRemoteWs = path == "/api/remote";
         var isRemoteScript = path == "/remote.ps1";
-        if (!path.StartsWith("/api/") || path == "/api/auth/login" || isTelegramWebhook || isSharedGet || isAutologin || isImageFile || isRemoteWs || isRemoteScript)
+        var isInternal = path.StartsWith("/api/internal/");
+        if (!path.StartsWith("/api/") || path == "/api/auth/login" || isTelegramWebhook || isSharedGet || isAutologin || isImageFile || isRemoteWs || isRemoteScript || isInternal)
         {
             await _next(ctx);
             return;

Plans/MCP_REMOTE_EXECUTE.md

@@ -0,0 +1,126 @@
+# Plan: MCP-Server für `remote_execute` in der CLI
+
+## Übersicht
+
+Ein eigener MCP-Server (Model Context Protocol, stdio-basiert) der `remote_execute` als Tool für die Claude Code CLI exposed. Die CLI erkennt MCP-Tools nativ und kann sie wie Built-in-Tools verwenden.
+
+## Architektur
+
+```
+Claude CLI ←→ MCP-Server (stdio) ←→ gott HTTP-API ←→ RemoteEndpoint → SSH/Exec
+```
+
+## Komponenten
+
+### 1. MCP-Server: `scripts/mcp-server` (oder `Services/McpServer.cs`)
+
+**Variante A: Standalone Script (Python/Node/Bash+jq)**
+- Leichtgewichtig, unabhängig von .NET Runtime
+- Kommuniziert über stdin/stdout (JSON-RPC 2.0)
+- Ruft gott's interne HTTP-API auf
+
+**Variante B: .NET Console App**
+- Kann ToolService/RemoteService direkt referenzieren
+- Gleiche Codebase, keine API-Zwischenschicht nötig
+- Aber: Eigener Prozess, muss separat gebaut werden
+
+**Empfehlung: Variante A** – ein einfaches Script das die gott-API aufruft. Minimaler Overhead, keine Build-Abhängigkeit.
+
+### 2. Interner API-Endpoint
+
+```
+POST /api/internal/remote-execute
+Authorization: (interner Token oder localhost-only)
+Content-Type: application/json
+
+{
+  "command": "docker ps",
+  "sessionId": "abc123",       // für Zuordnung zur Chat-Session
+  "serverId": "optional"       // falls mehrere Server
+}
+```
+
+**Response:**
+```json
+{
+  "exitCode": 0,
+  "stdout": "CONTAINER ID  IMAGE  ...",
+  "stderr": "",
+  "durationMs": 234
+}
+```
+
+### 3. MCP-Server Konfiguration
+
+In `.claude/settings.json` oder per `--mcp-config`:
+```json
+{
+  "mcpServers": {
+    "gott-remote": {
+      "command": "/app/gott/scripts/mcp-remote-execute",
+      "type": "stdio",
+      "env": {
+        "GOTT_API_URL": "http://localhost:5000",
+        "GOTT_SESSION_ID": "${SESSION_ID}"
+      }
+    }
+  }
+}
+```
+
+### 4. MCP-Protokoll (stdio, JSON-RPC 2.0)
+
+**Tool-Definition (tools/list Response):**
+```json
+{
+  "name": "remote_execute",
+  "description": "Führt einen Befehl auf dem Remote-Server aus (via SSH oder lokal)",
+  "inputSchema": {
+    "type": "object",
+    "properties": {
+      "command": {
+        "type": "string",
+        "description": "Der auszuführende Befehl"
+      }
+    },
+    "required": ["command"]
+  }
+}
+```
+
+**Tool-Call (tools/call):**
+- MCP-Server empfängt JSON-RPC Request
+- Ruft `POST /api/internal/remote-execute` auf
+- Gibt stdout/stderr als Text-Content zurück
+
+### 5. Integration in ClaudeCliService
+
+- MCP-Server-Config wird beim CLI-Start übergeben
+- `--mcp-config /tmp/gott-mcp-{sessionId}.json` oder globale Config
+- SessionId/ServerId werden als Environment-Variablen an den MCP-Server übergeben
+
+### 6. AnthropicService: CLI-Aufruf anpassen
+
+In `StreamChatViaCli` den MCP-Config-Pfad generieren und an `ChatStreamingAsync` übergeben:
+```csharp
+var mcpConfig = GenerateMcpConfig(sessionId, serverId);
+var mcpConfigPath = WriteTempMcpConfig(mcpConfig);
+// CLI-Aufruf mit --mcp-config mcpConfigPath
+```
+
+## Dateien
+
+| Datei | Aktion | Beschreibung |
+|-------|--------|-------------|
+| `scripts/mcp-remote-execute` | **NEU** | MCP-Server Script (stdio) |
+| `Endpoints/InternalEndpoint.cs` | **NEU** | Interner API-Endpoint für remote-execute |
+| `Services/ClaudeCliService.cs` | **EDIT** | MCP-Config übergeben |
+| `Services/AnthropicService.cs` | **EDIT** | MCP-Config generieren |
+| `.claude/settings.json` | **EDIT** | MCP-Server registrieren (oder dynamisch per --mcp-config) |
+
+## Offene Fragen
+
+1. **Script-Sprache**: Python (weit verbreitet), Node (npm-Ökosystem), oder Shell+curl+jq (keine Runtime-Abhängigkeit)?
+2. **Auth für internen Endpoint**: Nur localhost erlauben? Oder interner Token?
+3. **Server-Auswahl**: Wenn mehrere Remote-Server konfiguriert sind – soll der MCP-Server alle als separate Tools exposen oder einen Default verwenden?
+4. **Tool-Ergebnis-Mapping**: Wie sollen remote_execute-Results in der gott-UI angezeigt werden? Als normales Tool-Result oder speziell formatiert?

Program.cs

@@ -179,6 +179,7 @@ app.MapApiKeysEndpoints();
 app.MapOpenClawEndpoints();
 app.MapRemoteEndpoints();
 app.MapDatabaseExplorerEndpoints();
+app.MapInternalEndpoints();
 
 var port = app.Configuration.GetValue<int>("Port", 19000);
 app.Run($"http://0.0.0.0:{port}");

Services/AnthropicService.cs

@@ -1733,6 +1733,49 @@ public class AnthropicService
         }
     }
 
+    /// <summary>
+    /// Generiert eine temporäre MCP-Config-Datei für die Claude CLI.
+    /// Enthält den MCP-Server für remote_execute.
+    /// </summary>
+    private string GenerateMcpConfig(string sessionId)
+    {
+        // Port aus Konfiguration oder Default
+        var port = 19000;
+        var scriptPath = Path.Combine(AppContext.BaseDirectory, "scripts", "mcp-remote-execute.py");
+
+        // Fallback: Quell-Verzeichnis prüfen (für Dev-Modus)
+        if (!File.Exists(scriptPath))
+        {
+            // wwwroot ist typischerweise /app/gott/wwwroot → eine Ebene hoch = Projekt-Root
+            var projectRoot = Path.GetDirectoryName(AppPaths.WwwRoot) ?? AppContext.BaseDirectory;
+            var srcPath = Path.Combine(projectRoot, "scripts", "mcp-remote-execute.py");
+            if (File.Exists(srcPath)) scriptPath = srcPath;
+        }
+
+        var config = new
+        {
+            mcpServers = new Dictionary<string, object>
+            {
+                ["gott-remote"] = new
+                {
+                    command = "python3",
+                    args = new[] { scriptPath },
+                    env = new Dictionary<string, string>
+                    {
+                        ["GOTT_API_URL"] = $"http://localhost:{port}",
+                        ["GOTT_SESSION_ID"] = sessionId,
+                        ["GOTT_AUTH_TOKEN"] = Endpoints.InternalEndpoint.InternalToken
+                    }
+                }
+            }
+        };
+
+        var json = System.Text.Json.JsonSerializer.Serialize(config, new System.Text.Json.JsonSerializerOptions { WriteIndented = true });
+        var tempPath = Path.Combine(Path.GetTempPath(), $"gott-mcp-{sessionId}.json");
+        File.WriteAllText(tempPath, json);
+        return tempPath;
+    }
+
     /// <summary>
     /// Extrahiert Web-Search-Sources aus einem web_search_tool_result-Block.
     /// </summary>
@@ -1886,6 +1929,7 @@ public class AnthropicService
         bool planMode = false,
         bool webSearch = false,
         string sessionId = null,
+        RemoteConnectionManager remoteMgr = null,
         CancellationToken workCt = default,
         CancellationToken sseCt = default)
     {
@@ -1900,6 +1944,7 @@ public class AnthropicService
         var usage = new TokenUsage();
         var fullText = new StringBuilder();
         var tempImagePaths = new List<string>();
+        string mcpConfigPath = null;
 
         var chain = new ChainRecord
         {
@@ -1910,6 +1955,12 @@ public class AnthropicService
 
         try
         {
+            // MCP-Config für remote_execute generieren (wenn Remote-Verbindung vorhanden)
+            if (remoteMgr != null && !string.IsNullOrEmpty(sessionId) && remoteMgr.GetBySessionId(sessionId) != null)
+            {
+                mcpConfigPath = GenerateMcpConfig(sessionId);
+            }
+
             // System-Prompt als Text extrahieren
             var systemText = new StringBuilder();
             if (systemPrompt != null)
@@ -2079,6 +2130,7 @@ Danach STOPPE die Textgenerierung sofort. Nutze dies sparsam — nur wenn die En
                 hasWorkspace: !string.IsNullOrEmpty(workspace),
                 planMode: planMode,
                 webSearch: webSearch,
+                mcpConfigPath: mcpConfigPath,
                 cancellationToken: workCt
             );
 
@@ -2181,6 +2233,12 @@ Danach STOPPE die Textgenerierung sofort. Nutze dies sparsam — nur wenn die En
         {
             // Temp-Bilddateien aufräumen
             CleanupTempImages(tempImagePaths);
+
+            // MCP-Config Temp-Datei aufräumen
+            if (mcpConfigPath != null)
+            {
+                try { File.Delete(mcpConfigPath); } catch { }
+            }
         }
     }
 
@@ -2200,7 +2258,8 @@ Danach STOPPE die Textgenerierung sofort. Nutze dies sparsam — nur wenn die En
             "WebSearch" => "web_search",
             "WebFetch" => "web_fetch",
             "NotebookEdit" => "notebook_edit",
-            _ => cliName
+            "mcp__gott-remote__remote_execute" => "remote_execute",
+            _ => cliName?.StartsWith("mcp__") == true ? cliName.Split("__").LastOrDefault() ?? cliName : cliName
         };
     }
 

Services/ClaudeCliService.cs

@@ -82,6 +82,7 @@ public class ClaudeCliService
         bool hasWorkspace = false,
         bool planMode = false,
         bool webSearch = false,
+        string mcpConfigPath = null,
         CancellationToken cancellationToken = default)
     {
         if (!_isAvailable)
@@ -111,6 +112,8 @@ public class ClaudeCliService
             tools.AddRange(new[] { "Bash(*)", "Read", "Write", "Edit", "Glob", "Grep" });
         }
         if (webSearch) { tools.Add("WebFetch"); tools.Add("WebSearch"); }
+        // MCP-Tools erlauben (wenn MCP-Config vorhanden)
+        if (!string.IsNullOrEmpty(mcpConfigPath)) { tools.Add("mcp__gott-remote__remote_execute"); }
         args.Add("--allowed-tools");
         args.Add(tools.Count > 0 ? string.Join(" ", tools) : "\"\"");
 
@@ -132,6 +135,12 @@ public class ClaudeCliService
             args.Add(systemPrompt);
         }
 
+        if (!string.IsNullOrEmpty(mcpConfigPath))
+        {
+            args.Add("--mcp-config");
+            args.Add(mcpConfigPath);
+        }
+
         return await RunCliStreamingAsync(args, prompt, onEvent, cancellationToken);
     }
 

gott.csproj

@@ -47,6 +47,10 @@
     <Content Include="Templates/**" CopyToOutputDirectory="PreserveNewest" CopyToPublishDirectory="PreserveNewest" />
   </ItemGroup>
 
+  <ItemGroup>
+    <Content Include="scripts/**" CopyToOutputDirectory="PreserveNewest" CopyToPublishDirectory="PreserveNewest" />
+  </ItemGroup>
+
   <ItemGroup>
     <PackageReference Include="Cronos" Version="0.11.1" />
     <PackageReference Include="Microsoft.Extensions.Hosting.WindowsServices" Version="10.0.3" />

scripts/mcp-remote-execute.py

@@ -0,0 +1,217 @@
+#!/usr/bin/env python3
+"""
+MCP-Server (Model Context Protocol) für remote_execute.
+Kommuniziert über stdin/stdout (JSON-RPC 2.0) mit der Claude CLI.
+Ruft die gott-interne HTTP-API auf um Befehle auf dem Remote-PC auszuführen.
+
+Umgebungsvariablen:
+  GOTT_API_URL    - URL der gott-API (z.B. http://localhost:19000)
+  GOTT_SESSION_ID - Session-ID für die Remote-Verbindungszuordnung
+  GOTT_AUTH_TOKEN - Auth-Token für die gott-API (optional, für nicht-localhost)
+"""
+
+import json
+import sys
+import urllib.request
+import urllib.error
+import os
+
+API_URL = os.environ.get("GOTT_API_URL", "http://localhost:19000")
+SESSION_ID = os.environ.get("GOTT_SESSION_ID", "")
+AUTH_TOKEN = os.environ.get("GOTT_AUTH_TOKEN", "")
+
+TOOL_NAME = "remote_execute"
+TOOL_DESCRIPTION = (
+    "Führt einen PowerShell-Befehl auf der verbundenen Remote-Windows-Workstation aus. "
+    "Gibt stdout und stderr zurück. Nutze dies für Diagnose und Administration des Remote-PCs "
+    "(z.B. Druckerprobleme, Netzwerk, Services, Registry). Timeout: 180 Sekunden."
+)
+
+
+def send_response(response):
+    """Sendet eine JSON-RPC Response über stdout."""
+    data = json.dumps(response)
+    sys.stdout.write(data + "\n")
+    sys.stdout.flush()
+
+
+def handle_initialize(req_id, params):
+    """MCP initialize handshake."""
+    return {
+        "jsonrpc": "2.0",
+        "id": req_id,
+        "result": {
+            "protocolVersion": "2024-11-05",
+            "capabilities": {
+                "tools": {}
+            },
+            "serverInfo": {
+                "name": "gott-remote",
+                "version": "1.0.0"
+            }
+        }
+    }
+
+
+def handle_tools_list(req_id):
+    """Gibt die verfügbaren Tools zurück."""
+    return {
+        "jsonrpc": "2.0",
+        "id": req_id,
+        "result": {
+            "tools": [
+                {
+                    "name": TOOL_NAME,
+                    "description": TOOL_DESCRIPTION,
+                    "inputSchema": {
+                        "type": "object",
+                        "properties": {
+                            "command": {
+                                "type": "string",
+                                "description": "PowerShell-Befehl zum Ausführen auf dem Remote-PC"
+                            }
+                        },
+                        "required": ["command"]
+                    }
+                }
+            ]
+        }
+    }
+
+
+def handle_tools_call(req_id, params):
+    """Führt ein Tool aus."""
+    tool_name = params.get("name", "")
+    arguments = params.get("arguments", {})
+
+    if tool_name != TOOL_NAME:
+        return {
+            "jsonrpc": "2.0",
+            "id": req_id,
+            "result": {
+                "content": [{"type": "text", "text": f"Unbekanntes Tool: {tool_name}"}],
+                "isError": True
+            }
+        }
+
+    command = arguments.get("command", "")
+    if not command:
+        return {
+            "jsonrpc": "2.0",
+            "id": req_id,
+            "result": {
+                "content": [{"type": "text", "text": "Fehler: Kein Befehl angegeben"}],
+                "isError": True
+            }
+        }
+
+    # API-Call an gott
+    try:
+        url = f"{API_URL}/api/internal/remote-execute"
+        payload = json.dumps({
+            "command": command,
+            "sessionId": SESSION_ID
+        }).encode("utf-8")
+
+        req = urllib.request.Request(url, data=payload, method="POST")
+        req.add_header("Content-Type", "application/json")
+        if AUTH_TOKEN:
+            req.add_header("X-Internal-Token", AUTH_TOKEN)
+
+        with urllib.request.urlopen(req, timeout=200) as resp:
+            result = json.loads(resp.read().decode("utf-8"))
+
+        # Ergebnis formatieren
+        output_parts = []
+        if result.get("stdout"):
+            output_parts.append(result["stdout"])
+        if result.get("stderr"):
+            output_parts.append(f"[stderr] {result['stderr']}")
+        if result.get("exitCode", 0) != 0:
+            output_parts.append(f"[Exit Code: {result['exitCode']}]")
+
+        text = "\n".join(output_parts) if output_parts else "(keine Ausgabe)"
+        is_error = result.get("exitCode", 0) != 0
+
+        return {
+            "jsonrpc": "2.0",
+            "id": req_id,
+            "result": {
+                "content": [{"type": "text", "text": text}],
+                "isError": is_error
+            }
+        }
+
+    except urllib.error.HTTPError as e:
+        body = ""
+        try:
+            body = e.read().decode("utf-8")
+        except Exception:
+            pass
+        return {
+            "jsonrpc": "2.0",
+            "id": req_id,
+            "result": {
+                "content": [{"type": "text", "text": f"API-Fehler {e.code}: {body}"}],
+                "isError": True
+            }
+        }
+    except Exception as e:
+        return {
+            "jsonrpc": "2.0",
+            "id": req_id,
+            "result": {
+                "content": [{"type": "text", "text": f"Fehler: {str(e)}"}],
+                "isError": True
+            }
+        }
+
+
+def handle_notifications_initialized(params):
+    """Handle notifications/initialized - no response needed."""
+    pass
+
+
+def main():
+    """Hauptloop: Liest JSON-RPC Requests von stdin, verarbeitet sie."""
+    for line in sys.stdin:
+        line = line.strip()
+        if not line:
+            continue
+
+        try:
+            request = json.loads(line)
+        except json.JSONDecodeError:
+            continue
+
+        method = request.get("method", "")
+        req_id = request.get("id")
+        params = request.get("params", {})
+
+        # Notifications (kein id) brauchen keine Response
+        if req_id is None:
+            if method == "notifications/initialized":
+                handle_notifications_initialized(params)
+            continue
+
+        if method == "initialize":
+            response = handle_initialize(req_id, params)
+        elif method == "tools/list":
+            response = handle_tools_list(req_id)
+        elif method == "tools/call":
+            response = handle_tools_call(req_id, params)
+        else:
+            response = {
+                "jsonrpc": "2.0",
+                "id": req_id,
+                "error": {
+                    "code": -32601,
+                    "message": f"Method not found: {method}"
+                }
+            }
+
+        send_response(response)
+
+
+if __name__ == "__main__":
+    main()

wwwroot/css/chat.css

@@ -1280,6 +1280,14 @@
   color: var(--muted);
 }
 
+.inline-choice-block {
+  margin: var(--gap) 0;
+  padding: var(--gap);
+  border-left: 3px solid var(--primary);
+  background: var(--f1-bg);
+  border-radius: var(--border-radius);
+}
+
 /* ── OpenClaw Import Badge ── */
 
 .openclaw-badge {

wwwroot/js/chat.js

@@ -402,6 +402,14 @@ async function sendMessage() {
   await _runStream((signal) => apiChat(content, getCurrentSessionId(), signal, _planModeActive));
 }
 
+// ── Direct Message (für Inline-Choice-Buttons) ──
+async function sendMessageDirect(text) {
+  if (isStreaming || !text) return;
+  addUserBubble(text, true);
+  scrollDownForce();
+  await _runStream((signal) => apiChat(text, getCurrentSessionId(), signal, _planModeActive));
+}
+
 // ── Choice Click Handler ──
 
 async function handleChoiceClick(toolCallId, choice, bubbleDiv) {
@@ -419,3 +427,23 @@ async function handleChoiceClick(toolCallId, choice, bubbleDiv) {
 
   await _runStream((signal) => apiChatChoice(getCurrentSessionId(), toolCallId, choice, signal, _planModeActive));
 }
+
+// ── Inline Choice Blocks (Fallback aus Markdown-Rendering) ──
+// Event-Delegation für .inline-choice-btn Buttons die von renderMarkdown() erzeugt wurden
+document.addEventListener('click', function(e) {
+  var btn = e.target.closest('.inline-choice-btn');
+  if (!btn) return;
+  var block = btn.closest('.inline-choice-block');
+  if (!block || block.classList.contains('choice-answered')) return;
+
+  var choice = btn.dataset.choice;
+  // Buttons deaktivieren + markieren
+  block.querySelectorAll('.inline-choice-btn').forEach(function(b) {
+    b.disabled = true;
+    b.classList.add(b === btn ? 'choice-selected' : 'choice-unselected');
+  });
+  block.classList.add('choice-answered');
+
+  // Inline-Choice als normale User-Nachricht senden (kein tool_use in DB vorhanden)
+  sendMessageDirect(choice);
+});

wwwroot/js/markdown.js

@@ -6,6 +6,18 @@ function renderMarkdown(text) {
   if (!text) return '';
   var src = text.trim();
 
+  // 0) :::ASK_USER_CHOICE::: Blöcke erkennen und als Platzhalter speichern (vor Markdown-Parsing)
+  var choiceBlocks = [];
+  src = src.replace(/:::ASK_USER_CHOICE:::\s*\n?([\s\S]*?)\n?\s*:::END_ASK_USER_CHOICE:::/g, function(match, jsonStr) {
+    try {
+      var data = JSON.parse(jsonStr.trim());
+      choiceBlocks.push(data);
+      return '%%CHOICE' + (choiceBlocks.length - 1) + '%%';
+    } catch(e) {
+      return match; // JSON-Fehler → als Text belassen
+    }
+  });
+
   // 1) LaTeX-Ausdrücke VOR dem Markdown-Parsing schützen
   //    Reihenfolge: zuerst Block ($$...$$), dann Inline ($...$)
   var mathBlocks = [];
@@ -95,6 +107,22 @@ function renderMarkdown(text) {
     }
   });
 
+  // 8) Choice-Platzhalter durch interaktive Buttons ersetzen
+  html = html.replace(/%%CHOICE(\d+)%%/g, function(m, idx) {
+    var data = choiceBlocks[parseInt(idx)];
+    var q = data.question || '';
+    var choices = data.choices || [];
+    var jsonAttr = esc(JSON.stringify(data));
+    var out = '<div class="inline-choice-block" data-choice-json="' + jsonAttr + '">';
+    out += '<div class="choice-question">' + esc(q) + '</div>';
+    out += '<div class="choice-buttons">';
+    for (var i = 0; i < choices.length; i++) {
+      out += '<button class="choice-btn inline-choice-btn" data-choice="' + esc(choices[i]) + '">' + esc(choices[i]) + '</button>';
+    }
+    out += '</div></div>';
+    return out;
+  });
+
   return html;
 }
 

wwwroot/js/stream-ui.js

@@ -15,7 +15,16 @@ function startStreamBubble() {
 function appendToStream(text) {
   if (!_streamDiv) startStreamBubble();
   _streamText += text;
-  _streamDiv.innerHTML = renderMarkdown(_streamText);
+
+  // Wenn ein :::ASK_USER_CHOICE::: Block begonnen aber noch nicht abgeschlossen wurde,
+  // nur den Text VOR dem Marker rendern (Marker-Block verstecken bis er komplett ist)
+  var displayText = _streamText;
+  var choiceStart = displayText.indexOf(':::ASK_USER_CHOICE:::');
+  if (choiceStart >= 0 && displayText.indexOf(':::END_ASK_USER_CHOICE:::') < 0) {
+    displayText = displayText.substring(0, choiceStart);
+  }
+
+  _streamDiv.innerHTML = renderMarkdown(displayText);
   // Cursor ins letzte Inline/Text-Element einfügen, damit er nicht auf eine neue Zeile rutscht
   const cursor = document.createElement('span');
   cursor.className = 'cursor';

wwwroot/js/tool-ui.js

@@ -10,12 +10,19 @@ const CLI_TOOL_NAME_MAP = {
   'WebSearch': 'web_search',
   'WebFetch': 'web_fetch',
   'NotebookEdit': 'notebook_edit',
-  'ToolSearch': 'tool_search'
+  'ToolSearch': 'tool_search',
+  'mcp__gott-remote__remote_execute': 'remote_execute'
 };
 
 /** Normalize CLI tool name to gott convention (Edit → edit_file, etc.) */
 function normalizeToolName(name) {
-  return CLI_TOOL_NAME_MAP[name] || name;
+  if (CLI_TOOL_NAME_MAP[name]) return CLI_TOOL_NAME_MAP[name];
+  // Generisch: mcp__server__tool → tool
+  if (name && name.startsWith('mcp__')) {
+    var parts = name.split('__');
+    return parts[parts.length - 1] || name;
+  }
+  return name;
 }
 
 /** Normalize CLI tool input fields (file_path → path, old_string → old_text, etc.) */