Cursor Marketplace Served a Malicious Claude Code Extension With an SSH Backdoor

Cursor says that on October 29, 2026, between about 2:27 a.m. and 1:35 p.m. PT, malicious Anthropic.claude-code versions 2.1.312 and 2.1.313 were available in its VS Code extension marketplace and attempted to install a persistent SSH backdoor. The official Anthropic builds were not the source. Cursor pulled the versions and shut the marketplace down while it rebuilt it.

Christopher Ort
By Christopher Ort · 3 min read
Share
A quiet office kitchen at midday, a laptop closed on a counter, blinds open to a courtyard, no brand marks.
On this page
  1. Quick Take
  2. What the malware tried to do
  3. What you should do
  4. What it means
  5. About the analysis
  6. Sources

For about eleven hours on October 29, 2026, Cursor’s own extension marketplace handed developers a trojan. In a security notice published that day and updated on October 30, Cursor said “a maliciously modified version of Anthropic’s Claude Code extension” — Anthropic.claude-code versions 2.1.312 and 2.1.313 for Windows and macOS — was available in the Cursor VS Code Extension Marketplace from about 2:27 a.m. to 1:35 p.m. PT. “These malicious versions of the extension were not developed by Anthropic.”

i10X Quick Take

Quick Take

  1. What: Tampered Claude Code extension builds 2.1.312 and 2.1.313, served from Cursor’s marketplace, not from Anthropic’s official distribution.
  2. Window: About 2:27 a.m. to 1:35 p.m. PT on October 29, 2026.
  3. Payload: Collects host and user details and tries to install a persistent SSH backdoor, including an authorized-keys entry labeled svc-telemetry, a LaunchAgent or scheduled task, and a reverse tunnel to an attacker server. Cursor later had that server taken offline.
  4. Scale: The Verge reported that Cursor said around 5,000 people may have installed the malicious extension.
  5. Check: Search ~/.ssh/authorized_keys (and the Windows equivalents) for svc-telemetry. Uninstalling the extension is not enough; the payload drops standalone persistence.

What the malware tried to do

Cursor’s notice is unusually detailed. The extension collected hostname, username, OS, architecture, home directory, environment-variable names, working directories, configured Git email, and public IP, and reported them to the attacker. It tried to add an attacker-controlled SSH key, enable the local SSH service, and open a reverse SSH tunnel, then persist that tunnel with a macOS LaunchAgent (com.apple.sysmond) or a Windows scheduled task named SysMon, retrying about every 60 seconds.

Cursor also says the most damaging step often failed: turning on the local SSH service needs administrator rights, which the extension does not have on most Mac and Windows hosts. “On those hosts the step fails and the attacker could not have logged in through the tunnel.” That is not a clean bill of health. Cursor is clear that the extension still added the attacker’s SSH key, installed persistence, and sent host details, and it tells everyone in the window to follow the remediation steps anyway.

Cursor removed the builds, shut the marketplace down “out of an abundance of caution,” audited other extensions (it says it found no other tampering), rotated keys, rebuilt the marketplace, and worked with the attacker’s hosting provider to take the command-and-control server offline. Official Anthropic distribution and, by later reporting, other marketplaces were not the source of these builds.

What you should do

If you installed or updated Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor marketplace in that window:

  • Run Cursor’s check for the svc-telemetry key, then its remediation scripts (macOS as the affected user, Windows as Administrator). The scripts remove the extension, persistence, tunnel processes, and the key.
  • Isolate any host with evidence of payload execution. Uninstalling the extension alone may leave ~/.config/.smt, LaunchAgents, scheduled tasks, or startup scripts in place.
  • Rotate credentials and sessions that could be reached from an affected host, from a clean machine.

Cursor published indicators of compromise, including VSIX hashes for both versions and platforms, the attacker address 134.209.82.225, and the svc-telemetry key fingerprint. Microsoft October 30 analysis covers the same incident in more technical depth.

What it means

This was a marketplace compromise, not a bug in Claude Code’s upstream source. For developers, the lesson is brutal and familiar: the channel you use to install an official-looking editor extension is part of your trusted computing base. For Cursor, the eleven-hour window and the reported affected-user count make this one of the most serious AI-tool supply-chain incidents of the autumn. Even where the reverse-login step failed, key theft and persistence are enough to treat a hit machine as an incident, not a close call.

About the analysis

Researched and written with AI assistance under the editorial responsibility of the i10X News team. The title image is an AI-generated illustration. Read our AI Use Policy. Spotted an error? Request a correction.

This i10x report is based on Cursor’s October 29–30 security notice, The Verge’s report of Cursor’s affected-user figure, and Microsoft’s October 30 technical blog. i10x did not analyze the VSIX files.

Sources

Christopher Ort

Christopher Ort

Christopher Ort handles SEO, AI agent workflows and PR & partnerships at i10x.ai, the all-in-one AI workspace built by i10X Global Pte. Ltd. in Singapore. He is a graduate of Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). He is the author of “The Style Penalty”, a study published on SSRN that analyses 1,576 evaluation decisions by four frontier AI models to show how AI-written CVs fare in automated hiring screens; the research has been covered by heise online and t3n.

Timeline

What happened before

  1. Release

    Claude Haiku 5.5

    1M context · $0.1 / $0.5 per 1M tokens

    support.claude.com (source, opens in a new tab)
  2. Release

    Claude Sonnet 5.5

    1M context · $2 / $10 per 1M tokens

    support.claude.com (source, opens in a new tab)
  3. Release

    Claude Opus 5.5

    1M context · $4 / $20 per 1M tokens

    First model of the Claude 5.5 family; performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5.

    anthropic.com (source, opens in a new tab)
  4. Release

    Claude Fable 5.1 & Mythos 5.1

    1M context · $10 / $50 per 1M tokens

    Anthropic's most advanced models for coding and knowledge work. Price/context shown for Claude Fable 5.1 (models overview).

    support.claude.com (source, opens in a new tab)
  5. Release

    Claude Opus 5

    1M context · $5 / $25 per 1M tokens

    support.claude.com (source, opens in a new tab)
Full Claude timeline ›

Editor's picks

  1. CVE-2026-77604: Path Traversal Vulnerability in Code Graph RAG MCP Server Allows Arbitrary File Read
  2. CVE-2026-123271: ChatCLI OAuth Flaw Allows Token Theft via Malicious MCP Servers
  3. CVE-2026-123022: AgentScope MCP Workspace Allows Unauthenticated Remote Code Execution
  4. CVE-2026-97677: Path Traversal Flaw in IBM Langflow OSS 1.0.0–1.12.2 Allows File Read/Write
  5. Active Support DoS Flaw Lets Unauthenticated Attackers Crash Rails Servers via Action Cable