@0xmaxma/claude-gateway 1.2.12 → 1.2.15

package/mcp/tools/api/module.ts

@@ -0,0 +1,87 @@
+import type { ToolModule, McpToolDefinition, McpToolResult, ToolVisibility } from '../../types';
+
+export class ApiModule implements ToolModule {
+  id = 'api';
+  toolVisibility: ToolVisibility = 'current-channel';
+
+  isEnabled(): boolean {
+    return process.env.GATEWAY_ORIGIN_CHANNEL === 'api';
+  }
+
+  getTools(): McpToolDefinition[] {
+    return [
+      {
+        name: 'api_reply',
+        description:
+          'Attach files to the current API session response. ' +
+          'Use to include screenshots or images in the reply returned to the API caller. ' +
+          'Files must be absolute paths already saved to the session media directory.',
+        inputSchema: {
+          type: 'object',
+          properties: {
+            files: {
+              type: 'array',
+              items: { type: 'string' },
+              description: 'Absolute file paths to attach as images.',
+            },
+          },
+          required: ['files'],
+        },
+      },
+    ];
+  }
+
+  async handleTool(name: string, args: Record<string, unknown>): Promise<McpToolResult> {
+    if (name === 'api_reply') {
+      return this.handleApiReply(args);
+    }
+    return { content: [{ type: 'text', text: `Unknown tool: ${name}` }], isError: true };
+  }
+
+  private async handleApiReply(args: Record<string, unknown>): Promise<McpToolResult> {
+    const files = (args.files as string[] | undefined) ?? [];
+    if (!files.length) {
+      return { content: [{ type: 'text', text: 'No files provided.' }] };
+    }
+
+    const apiUrl = process.env.GATEWAY_API_URL;
+    const agentId = process.env.GATEWAY_AGENT_ID;
+    const sessionId = process.env.GATEWAY_SESSION_ID;
+    const apiKey = process.env.GATEWAY_API_KEY;
+
+    if (!apiUrl || !agentId || !sessionId) {
+      return {
+        content: [{ type: 'text', text: 'api_reply: missing GATEWAY_API_URL, GATEWAY_AGENT_ID, or GATEWAY_SESSION_ID' }],
+        isError: true,
+      };
+    }
+
+    try {
+      const res = await fetch(
+        `${apiUrl}/v1/agents/${encodeURIComponent(agentId)}/sessions/${encodeURIComponent(sessionId)}/attachments`,
+        {
+          method: 'POST',
+          headers: {
+            'Content-Type': 'application/json',
+            ...(apiKey ? { 'X-Api-Key': apiKey } : {}),
+          },
+          body: JSON.stringify({ files }),
+          signal: AbortSignal.timeout(5000),
+        },
+      );
+      if (!res.ok) {
+        const text = await res.text().catch(() => '');
+        return {
+          content: [{ type: 'text', text: `api_reply: gateway returned ${res.status}: ${text}` }],
+          isError: true,
+        };
+      }
+      return { content: [{ type: 'text', text: `Attached ${files.length} file(s).` }] };
+    } catch (err) {
+      return {
+        content: [{ type: 'text', text: `api_reply: ${(err as Error).message}` }],
+        isError: true,
+      };
+    }
+  }
+}

package/mcp/tools/browser/module.ts

@@ -24,16 +24,37 @@ export class BrowserModule implements ToolModule {
 
     const result = await callGetpodBrowser(name, args);
 
+    if (name === 'browser_create_session' && !result.isError) {
+      const textBlock = result.content[0] as { type: string; text: string } | undefined;
+      if (textBlock?.type === 'text') {
+        try {
+          const parsed = JSON.parse(textBlock.text) as Record<string, unknown>;
+          delete parsed['stream_url'];
+          return { content: [{ type: 'text', text: JSON.stringify(parsed) }] };
+        } catch {
+          // return as-is if parsing fails
+        }
+      }
+    }
+
     if (name === 'browser_screenshot' && !result.isError) {
       // getpod-browser returns {type:"image", data: base64, mimeType:"image/jpeg"}.
-      // Decode and save to /tmp so callers can attach the file path directly.
+      // Decode and save so callers can attach the file path directly.
+      // For API sessions, save to the session media dir so files are HTTP-accessible.
       const block = result.content[0] as Record<string, string> | undefined;
       const b64 = block?.['data'] ?? '';
       const mime = block?.['mimeType'] ?? 'image/jpeg';
       if (b64) {
         const ext = mime.includes('png') ? 'png' : 'jpg';
         const sid = (args.session_id as string | undefined) ?? 'default';
-        const filePath = path.join('/tmp', `browser_shot_${sid}.${ext}`);
+        const mediaDir = process.env.GATEWAY_SESSION_MEDIA_DIR;
+        let filePath: string;
+        if (mediaDir) {
+          fs.mkdirSync(mediaDir, { recursive: true });
+          filePath = path.join(mediaDir, `browser_shot_${sid}_${Date.now()}.${ext}`);
+        } else {
+          filePath = path.join('/tmp', `browser_shot_${sid}.${ext}`);
+        }
         fs.writeFileSync(filePath, Buffer.from(b64, 'base64'));
         return { content: [{ type: 'text', text: filePath }] };
       }
@@ -136,9 +157,7 @@ async function callGetpodBrowser(
 const browserToolDefs: McpToolDefinition[] = [
   {
     name: 'browser_create_session',
-    description:
-      'Create or resume a browser session. Returns stream_url and status. ' +
-      'IMPORTANT: After creating a session, always share the stream_url with the user so they can open the browser in their client.',
+    description: 'Create or resume a browser session. Returns session status.',
     inputSchema: {
       type: 'object',
       properties: {

package/mcp/tools/browser/skills/open-browser/SKILL.md

@@ -1,12 +1,12 @@
 ---
 name: open-browser
-description: "Open a URL in the agent's persistent browser session via MCP browser tools"
+description: "ALWAYS invoke this skill when user says 'browser [site]', 'open [site]', or asks to navigate to a website. Never call MCP browser tools directly."
 user-invocable: true
 ---
 
 # open-browser
 
-When user says "เปิด X", "navigate to X", "open X in browser", "switch to tab X", etc.
+When user says "open X in browser", "navigate to X", "open chrome", "browser to X", "switch to tab X", etc.
 
 ## Rules
 
@@ -21,13 +21,9 @@ When user says "เปิด X", "navigate to X", "open X in browser", "switch t
 
 Call `mcp__gateway__browser_create_session` with NO arguments (session_id is auto-injected).
 
-Result contains `stream_url` — use it to build the browser URL for the user.
+Result contains session status only.
 
-### Step 2 — Send browser URL to user IMMEDIATELY (no waiting)
-
-Reply on Telegram right away with the stream_url so the user can open the browser.
-
-### Step 3 — Check current tabs
+### Step 2 — Check current tabs
 
 Call `mcp__gateway__browser_tabs` — returns list of `{tab_id, url, title}`.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@0xmaxma/claude-gateway",
-  "version": "1.2.12",
+  "version": "1.2.15",
   "description": "Multi-agent gateway for Claude",
   "repository": {
     "type": "git",