AAK-MCP-LANGBOT-CVE-2026-54449-001: LangBot lets any authenticated user configure an STDIO MCP command (no fix)
- Severity
- HIGH
- Category
- Supply Chain (
supply-chain) - Scanner
agent_audit_kit/scanners/mcp_cve_pins_2026_07.py- Auto-fixable
- No.
- SARIF name
LangBotStdioMcpAuthzBypass
What it detects
LangBot at 4.10.7 and earlier applies no adequate authorization boundary to the Extensions MCP configuration: in src/langbot/pkg/provider/tools/loaders/mcp.py, StdioServerParameters takes the configured command and arguments and starts a server-side subprocess, so anyone who can sign up or obtain an account executes arbitrary commands with the privileges of the LangBot service (CVE-2026-54449, CVSS 3.1 8.8). This is a presence-only pin because no fix exists: GHSA-3pvh-63gf-j9mw records no patched version, NVD states none is available, and 4.10.8 -- published 31 minutes after the advisory -- is a media-delivery patch whose release notes claim no security fix. Treat every published version as exposed until upstream states otherwise.
How to fix
No upgrade resolves this, so do not wait for one. Close the account boundary instead: disable open sign-up, and restrict the Extensions MCP configuration surface to operators rather than any authenticated user. If the deployment cannot enforce that, take the STDIO MCP configuration path out of the reachable UI, since the subprocess runs with the service's own privileges. Re-pin to a version floor once upstream ships a release that states it fixes this.
To confirm the finding is gone, report only this rule:
agent-audit-kit scan . --rules AAK-MCP-LANGBOT-CVE-2026-54449-001
Limitations
Reads langbot in requirements*.txt at any depth and, at the project root only, in pyproject.toml, Pipfile, package.json, .mcp.json, mcp.json, claude_desktop_config.json and the Pipfile.lock, poetry.lock, uv.lock, package-lock.json, pnpm-lock.yaml and yarn.lock lockfiles, and reports every version, a future fixed one included, until the pin gets a floor. The dependency is not read when it is declared in a sub-directory manifest (services/api/pyproject.toml) or in a lockfile format the reader does not parse (pdm.lock). A compose file's rockchin/langbot image tag is not read. A file over 2 MB, or a lockfile over 32 MiB, is not read. A declaration nested more than 64 levels deep in package.json, a TOML manifest or an MCP config (an MCP server more than 62) is still judged, and reported without a line; a lockfile entry keeps its line at any depth.
References
- CVE: CVE-2026-54449
- OWASP MCP Top 10: MCP04:2025 (Software Supply Chain Attacks & Dependency Tampering), MCP05:2025 (Command Injection & Execution), MCP07:2025 (Insufficient Authentication & Authorization)
- OWASP Agentic Top 10: ASI03 (Identity and Privilege Abuse), ASI04 (Agentic Supply Chain Vulnerabilities), ASI05 (Unexpected Code Execution (RCE))
- Adversa AI MCP Top 25 (AAK-defined ids): ADV-INJECT-01 (Shell Injection via MCP)
- CSA AI Controls Matrix (AICM): IVS-04, IAM-01
All rules. This page is generated from the rule registry when the documentation is built; to change it, change the rule in agent_audit_kit/rules/builtin.py. To give the rule a hand-written write-up, add docs/rules/AAK-MCP-LANGBOT-CVE-2026-54449-001.md: the page then shows that text, with the registry reference appended under it.