CVE-2026-81735: UI-TARS-desktop @agent-infra MCP Servers Bind Every Interface Without Authentication, Exposing Arbitrary Command Execution
Advisory generated from the CVE record, CISA, Exploit Observer and VEDAS: what it is, whether it is exploited, who is affected, and how to fix and detect it.
Generated by VEDAS. Scores as of 2026-11-01 19:19 UTC. Rebuilt every few hours from cve-scores, CISA KEV and vedas-signatures. How this works and its limits.
What it is
startServer.ts in the mcp-http-server package of UI-TARS-desktop defaulted its listen address to ‘::’ when no host was given, so startSseAndStreamableHttpMcpServer bound the Streamable HTTP and SSE MCP transports to every interface, and its authentication middleware was optional: middlewares are applied only when a caller supplies them. The @agent-infra/mcp-server-commands and @agent-infra/mcp-server-filesystem entry points call startSseAndStreamableHttpMcpServer with a host and port alone and pass no middleware, so neither server required a credential. The commands server exposes a run_command tool that hands its caller-supplied command string to promisify(child_process.exec), so any unauthenticated client able to reach the port could run arbitrary commands as the user running the server, and the filesystem server exposed its file read and write tools on the same terms. The listen default became 127.0.0.1 in commit c2ad42e3eb9b27830db41a3e6f51ca7179d9b168; the package version stayed at 1.2.4 across that change, so the boundary is the commit rather than a release.
Description quoted from the CVE record (assigned by VulnCheck, published 2026-08-27).
Weakness: Missing Authentication for Critical Function
How it can be exploited
Exploitable over the network with low attack complexity with no special attack requirements needing no privileges and no user interaction. Impact: confidentiality high, integrity high, availability high. CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H
CISA assessment (SSVC): CISA has no evidence of exploitation. Automatable: yes. Technical impact: total.
Public exploit evidence (Exploit Observer)
Also known as: EUVD-2026-67078, GHSA-p68p-fq8j-3639
VEDAS cluster: VEDAS:DEFTDYNAMO
Exploit Observer links 5 pieces of public exploit code or detection modules that name this CVE or one of its aliases.
Exploit code and modules (5)
Code that names this CVE. A repository can be a working exploit, a scanner, a fake or malware; read it before running anything.
- Code repository: https://raw.githubusercontent.com/ARPSyndicate/vedas-signatures/main/nuclei/2026/CVE-2026-81735.yaml
- Code repository: https://raw.githubusercontent.com/ARPSyndicate/vedas-signatures/main/nuclei/2026/CVE-2026-81735.yaml
- Code repository: https://raw.githubusercontent.com/ARPSyndicate/vedas-signatures/main/suricata/2026/CVE-2026-81735.rules
- Code repository: https://raw.githubusercontent.com/ARPSyndicate/vedas-signatures/main/suricata/2026/CVE-2026-81735.rules
- Code repository: https://raw.githubusercontent.com/ARPSyndicate/vedas-signatures/main/suricata/2026/CVE-2026-81735.rules
Vulnerability databases and vendor advisories (3)
Catalogue entries, not exploit evidence.
- Vulnerability database or vendor advisory: https://euvd.enisa.europa.eu/enisa/EUVD-2026-67078
- Vulnerability database or vendor advisory: https://github.com/advisories/GHSA-p68p-fq8j-3639
- Vulnerability database or vendor advisory: https://nvd.nist.gov/vuln/detail/CVE-2026-81735
2 more references returned by Exploit Observer did not name this CVE or an alias, so they are left out.
VEDAS data as of 2026-11-01T17:12 UTC; fetched 2026-11-02.
Live scores
- EPSS
- 0.008
low predicted chance of attack - VEDAS
- 0.989
strong exploit evidence - On CISA KEV
- Not listed
Who is affected
| Product | Affected | Not affected |
|---|---|---|
| bytedance UI-TARS-desktop | before c2ad42e3eb9b (git commit) | c2ad42e3eb9b is not affected |
Version ranges exactly as the CVE record states them. “Before X” means X is the first fixed version.
How to fix it
Vendor guidance:
- Pull Request #1918 (issue-tracking, patch)
- Patch Commit (patch)
How to detect it
Nuclei template (active check)
UI-TARS-desktop mcp-http-server - Unauthenticated MCP Command Execution Exposure
The mcp-http-server package of UI-TARS-desktop defaulted its listen address to ‘::’ and applied authentication middleware only when a caller supplied one. The @agent-infra/mcp-server-commands and @agent-infra/mcp-server-filesystem entry points start the Streamable HTTP / SSE MCP transports with no middleware, so any unauthenticated client able to reach the port can complete the MCP handshake and invoke the run_command tool, which passes a caller-supplied command to child_process.exec, yielding remote command execution. Fixed by defaulting the listen address to 127.0.0.1 in commit c2ad42e3.
Name and description quoted from the template, written by VEDAS.
The template sends 1 request (POST /mcp); it decides using dsl matchers. Severity in the template: critical. Marked as not yet verified against a live vulnerable target.
Run it against a host you own or are authorised to test:
nuclei -u https://your-host -t CVE-2026-81735.yamlShow the template (42 lines)
id: CVE-2026-81735
info:
name: "UI-TARS-desktop mcp-http-server - Unauthenticated MCP Command Execution Exposure"
author: vedas-arpsyndicate
severity: critical
description: |
The mcp-http-server package of UI-TARS-desktop defaulted its listen address to '::' and
applied authentication middleware only when a caller supplied one. The
@agent-infra/mcp-server-commands and @agent-infra/mcp-server-filesystem entry points start
the Streamable HTTP / SSE MCP transports with no middleware, so any unauthenticated client
able to reach the port can complete the MCP handshake and invoke the run_command tool,
which passes a caller-supplied command to child_process.exec, yielding remote command
execution. Fixed by defaulting the listen address to 127.0.0.1 in commit c2ad42e3.
reference:
- https://www.exploit.observer/?keyword=CVE-2026-81735&match=exact
- https://subdomain.center/?engine=ammonites&keyword=ui-tars&match=exact
classification:
cvss-metrics: CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H
cvss-score: 10.0
cve-id: CVE-2026-81735
cwe-id: CWE-306
metadata:
verified: false
tags: cve,cve2026,ui-tars,mcp,rce
http:
- raw:
- |
POST /mcp HTTP/1.1
Host: {{Hostname}}
Content-Type: application/json
Accept: application/json, text/event-stream
{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"mcp","version":"1.0"}}}
matchers:
- type: dsl
dsl:
- 'contains(body, "protocolVersion")'
- 'contains(body, "jsonrpc")'
- 'status_code == 200'
condition: and
Suricata rules (network detection)
1 rule that alert on traffic matching this exploit: VEDAS CVE-2026-81735 UI-TARS-desktop mcp-http-server - Unauthenticated MCP Handshake Attempt. Add them to your Suricata rule set and reload.
Show the rules
alert http $EXTERNAL_NET any -> $HOME_NET any (msg:"VEDAS CVE-2026-81735 UI-TARS-desktop mcp-http-server - Unauthenticated MCP Handshake Attempt"; flow:established,to_server; http.method; content:"POST"; http.uri; content:"/mcp"; nocase; http.request_body; content:"initialize"; nocase; fast_pattern; content:"jsonrpc"; nocase; classtype:attempted-admin; reference:cve,CVE-2026-81735; reference:url,https://www.exploit.observer/?keyword=CVE-2026-81735&match=exact; reference:url,https://subdomain.center/?engine=ammonites&keyword=ui-tars&match=exact; sid:1081401; rev:1;)
CVE-2026-81735.rules on GitHub
References
Vendor advisories and patches
- Pull Request #1918 (issue-tracking, patch)
- Patch Commit (patch)
Exploits and technical analysis
- https://github.com/bytedance/UI-TARS-desktop/blob/e9f3387288da4af2ad99972da2ac916cdabce093/packages/agent-infra/mcp-http-server/src/startServer.ts#L263 (technical-description)
- VulnCheck Advisory: UI-TARS-desktop @agent-infra MCP Servers Bind Every Interface Without Authentication, Exposing Arbitrary Command Execution (third-party-advisory)
Other references
Other databases
Where this comes from
This advisory is generated automatically by VEDAS from: CVE.org record, cve-scores (EPSS and VEDAS), Exploit Observer, vedas-signatures. Text in quotes is reproduced from the source; everything else is a fixed template filled with source fields. Nothing here is written by a language model. Check the vendor advisory before acting, and read how VEDAS and Exploit Observer work, and their limits.