Cursor has published the results of its two-week investigation into the incident that put a fake Claude Code extension in its marketplace. In its update, Cursor says the attack began at Fireworks, a company that provides AI model inference for Cursor. Cursor had issued Fireworks an access key, Fireworks exposed it, and an attacker used it to access a read-only storage bucket the two companies used to share data.

The bucket held coding agent traces, which are records of agent sessions. Cursor says they were mostly from 2025 and only from sessions not in Privacy Mode, which Cursor was permitted to use for model training. A small number contained credentials that had not been redacted, and Cursor says less than 0.01% of customers had API credentials exposed.

How a trace file became a marketplace attack

Cursor says the attacker used those credentials to reach select SaaS tools Cursor uses. From there they accessed archived non-production source code repositories, limited customer support data and email conversations, product usage metrics and contact details, and summaries of some coding agent conversations. Cursor says some credentials users had sent to its support team were also accessed.

On October 29, Cursor says, the attacker used a stolen credential from that exposed data to place malicious copies of Anthropic’s Claude Code extension in the marketplace cache. Cursor’s update names versions 2.1.312 and 2.1.313. Cursor’s first notice, covered here on October 30, named only 2.1.313. Cursor says at most 5,000 users who requested those cached versions that day could have received the malicious build, and that it last detected and blocked attacker activity on November 1.

What to check

Cursor says it has contacted each affected customer directly and that the vast majority were not affected. If you installed Claude Code from the Cursor marketplace on October 29, follow the steps in Cursor’s security notice. Cursor warns that uninstalling the extension alone may not be enough.

Even if you got no message, two other exposures are worth checking:

  • Secrets you sent to support. If you ever pasted an API key or password into an email or chat with Cursor’s support team, rotate it. Cursor says some such credentials were accessed.
  • Secrets in agent sessions. If you pasted a key into an agent chat in 2025 with Privacy Mode off, treat it as possibly exposed and replace it. Cursor says it redacted credentials in these traces imperfectly.

Cursor says it has moved all inference traffic off Fireworks, rotated the exposed credentials and all credentials in use across its internal systems, added secret scanning across its code repositories, and redeployed all production infrastructure.


End of article