MCPoison is a remote code execution vulnerability in Cursor, an AI-assisted code editor, affecting versions 1.2.4 and earlier. The flaw arises from Cursor’s handling of previously trusted Model Context Protocol (MCP) server configurations: once a user has approved an MCP configuration, subsequent modifications to that trusted configuration are not re-validated or re-prompted before execution. An attacker who can modify an already approved MCP configuration in a shared source repository or write to the local configuration on the target system can replace a benign command with a malicious one, causing Cursor to execute attacker-controlled commands without additional warning. The issue has been described as binding trust to the approved MCP entry rather than to the integrity of the approved command content, enabling a silent swap of trusted configuration into malicious execution.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
No valid public exploits. Mallory filtered out 4 candidates as fakes, detection scripts, or README-only repos.
All candidate exploits were filtered out by Mallory's validation.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
31 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
An earlier Cursor vulnerability allowing attackers to abuse MCP configuration approval flow so malicious commands could be swapped in after initial approval without a second prompt.
A similar vulnerability in Cursor involving automatic execution of workspace configurations through MCP, mentioned for comparison only.
A vulnerability in Cursor where project-level MCP configuration could lead to command execution.
A related vulnerability mentioned as part of a broader pattern of auto-executing workspace configurations in AI coding tools; no technical details are provided in the content.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.