Loading vulnerability…
CVE-2026-5029 is a remote code execution vulnerability in Code Runner MCP Server when it is started with the --transport http option. In this configuration, the server exposes the /mcp JSON-RPC endpoint on port 3088 without authentication. An unauthenticated remote attacker can call the run-code MCP tool and provide arbitrary source code, which the server executes via child_process.exec() using the attacker-specified language interpreter. As a result, arbitrary commands or code can be executed in the security context of the user account running the server. Based on the provided content, the issue may affect all versions and no fix was available at the time of disclosure.
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.
1 valid exploit after Mallory filtered fakes, detection scripts, and README-only repos.
This repository is a small standalone Python exploit project containing two files: a README and a single executable script, exploit.py. The script is the main artifact and implements an unauthenticated remote code execution exploit for CVE-2026-5029 affecting Code Runner MCP Server when launched with HTTP transport enabled. The exploit is not just a detector: it is designed for active exploitation. Based on the README and visible code, it uses the fastmcp Client to connect to an MCP JSON-RPC endpoint, defaulting to http://localhost:3088/mcp unless overridden with -u/--url. It supports three main workflows: (1) interactive exploitation via an RCE shell, (2) enumeration of available languages/tools exposed by the server, and (3) probing which language runtimes are actually installed on the target. The default execution language is shellscript, but the CLI allows many language identifiers to be selected. Operationally, the exploit targets the vulnerable run-code MCP tool. The intended flow is to connect to /mcp, inspect available tools/schema, and then invoke tools/call with attacker-controlled code and a languageId so the server executes the supplied code through an interpreter. The provided result is arbitrary command/code execution with the privileges of the server process. Repository structure is simple and purpose-built: README.md documents the vulnerability, usage, and remediation; exploit.py contains the exploit logic and a Rich-based terminal UI for banners, status panels, and interactive output. No evidence suggests this belongs to a major exploit framework such as Metasploit or Nuclei. The maturity is best classified as OPERATIONAL because it includes a usable payload delivery mechanism and interactive shell behavior, but not a highly modular weaponized framework.
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.
2 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
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.