agent control layer: vision endpoint (qwen2.5vl:7b), tap/swipe/text/launch, fleet actions, onboard retry sweep, MCP server :8095

mcp_server.py

@@ -0,0 +1,136 @@
+#!/usr/bin/env python3
+"""
+Android Fleet MCP server — lets Hermes agents (and any MCP client) drive the
+fleet: see screens, tap, type, run commands, trigger onboarding, broadcast.
+HTTP streamable-http on :8095/mcp. All calls proxy to the godhead REST API.
+"""
+import json
+import urllib.request
+
+from mcp.server.fastmcp import FastMCP
+
+GODHEAD = "http://127.0.0.1:8080"
+
+mcp = FastMCP("android-fleet", host="0.0.0.0", port=8095)
+
+def _api(path, method="GET", payload=None, timeout=120):
+    data = json.dumps(payload).encode() if payload is not None else None
+    req = urllib.request.Request(GODHEAD + path, data=data, method=method,
+                                 headers={"Content-Type": "application/json"})
+    with urllib.request.urlopen(req, timeout=timeout) as r:
+        return json.loads(r.read())
+
+def _err(e):
+    return {"error": str(e)[:200]}
+
+@mcp.tool()
+def fleet_devices() -> dict:
+    """List all fleet devices with status, IP, proxy and IG account state."""
+    try:
+        devs = _api("/api/devices")
+        accts = {a["mac"]: a.get("status") for a in _api("/api/accounts")}
+        for d in devs:
+            d["ig_status"] = accts.get(d["mac"], "none")
+        return {"count": len(devs),
+                "online": sum(1 for d in devs if d["adb_state"] == "online"),
+                "devices": devs}
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def device_shell(mac: str, cmd: str) -> dict:
+    """Run an adb shell command on one device (mac = device MAC address)."""
+    try:
+        return _api(f"/api/device/{mac}/exec", "POST", {"cmd": cmd})
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def device_vision(mac: str, instruction: str) -> dict:
+    """Screenshot the device and ask the vision model (qwen2.5vl:7b on GPU)
+    about what's on screen. Use before tapping to find coordinates."""
+    try:
+        return _api(f"/api/device/{mac}/vision", "POST", {"instruction": instruction})
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def device_tap(mac: str, x: int, y: int) -> dict:
+    """Tap screen coordinates on a device."""
+    try:
+        return _api(f"/api/device/{mac}/tap", "POST", {"x": x, "y": y})
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def device_swipe(mac: str, x1: int, y1: int, x2: int, y2: int, ms: int = 300) -> dict:
+    """Swipe on a device (e.g. scroll the feed)."""
+    try:
+        return _api(f"/api/device/{mac}/swipe", "POST",
+                    {"x1": x1, "y1": y1, "x2": x2, "y2": y2, "ms": ms})
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def device_type(mac: str, text: str) -> dict:
+    """Type text on a device (into the focused field)."""
+    try:
+        return _api(f"/api/device/{mac}/text", "POST", {"text": text})
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def device_launch(mac: str, package: str) -> dict:
+    """Launch an app by package name on a device."""
+    try:
+        return _api(f"/api/device/{mac}/launch", "POST", {"package": package})
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def device_screenshot(mac: str) -> dict:
+    """Take a screenshot; returns a vision description + the PNG is cached
+    at /api/device/<mac>/screenshot/latest."""
+    try:
+        shot = _api(f"/api/device/{mac}/screenshot", "POST", timeout=60)
+        return {"saved": shot.get("path"), "view": f"/api/device/{mac}/screenshot/latest"}
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def fleet_broadcast_tap(macs: list[str], x: int, y: int) -> dict:
+    """Tap the same coordinates on many devices at once."""
+    try:
+        return _api("/api/fleet/action", "POST",
+                    {"macs": macs, "action": {"kind": "tap", "params": {"x": x, "y": y}}})
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def onboard_device(mac: str) -> dict:
+    """Run the full phone+IG onboarding on a device (creates IG account when
+    Instagram allows it; rate limits are reported honestly)."""
+    try:
+        return _api(f"/api/device/{mac}/onboard", "POST", {})
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def spawn_devices(count: int) -> dict:
+    """Clone new Android VMs from the template; they auto-register."""
+    try:
+        return _api("/api/spawn", "POST", {"count": count})
+    except Exception as e:
+        return _err(e)
+
+@mcp.tool()
+def fleet_message(text: str) -> dict:
+    """Broadcast a message via saved IG sessions (requires created accounts)."""
+    try:
+        return _api("/api/fleet/message", "POST", {"text": text})
+    except Exception as e:
+        return _err(e)
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(mcp.streamable_http_app(), host="0.0.0.0", port=8095, log_level="warning")