CVE-2026-45018: Unauthenticated Remote Code Execution via MCP stdio Transport in Chainlit
Vulnerability ID: CVE-2026-45018
CVSS Score: 9.8
Published: 2026-08-25
CVE-2026-45018 is a critical command injection vulnerability in Chainlit's Model Context Protocol (MCP) stdio transport backend. By submitting a crafted JSON payload containing dangerous argument options to an unauthenticated HTTP endpoint, a remote attacker can bypass executable validation rules and run arbitrary shell commands with the privileges of the active Python process.
TL;DR
Unauthenticated command injection vulnerability in Chainlit's /mcp endpoint allows remote attackers to execute arbitrary system commands via shell execution arguments inside the stdio transport parameters.
Technical Details
- CWE ID: CWE-78
- Attack Vector: Network
- CVSS Score: 9.8 (Critical)
- Exploit Status: Proof-of-Concept
- CISA KEV Status: Not Listed
- EPSS Score: Not Available
- Impact: Arbitrary Code Execution
Affected Systems
- Chainlit Applications
-
chainlit: >= 2.4.0rc0, < 2.12.0 (Fixed in:
2.12.0)
Code Analysis
Commit: 0565fd0
Refactor MCP integration to remove client-side command execution and enforce server-side static configurations
Mitigation Strategies
- Upgrade Chainlit to 2.12.0 or higher
- Disable Model Context Protocol by setting features.mcp.enabled = false in config.toml
- Restrict outbound network egress traffic from application servers
Remediation Steps:
- Check current installed package version using
pip show chainlit. - If version is within the vulnerability range, append
enabled = falseinside the[features.mcp]configuration block in.chainlit/config.toml. - Execute package upgrade command
pip install -U chainlit>=2.12.0. - Verify service startup behavior and validate that the
/mcppath rejects legacyfullCommandarguments.
References
- Chainlit Security Advisory
- Official Fix Commit
- Chainlit Release Note (v2.12.0)
- GitHub Security Advisory (GHSA)
Read the full report for CVE-2026-45018 on our website for more details including interactive diagrams and full exploit analysis.
Top comments (0)