Quick Take
- What happened: An attacker obtained a Fireworks AI access key that gave read access to a Cursor storage bucket holding coding agent traces, some with unredacted credentials.
- Who is affected: Under 0.05% of Cursor users not in Privacy Mode had traces exposed; under 0.01% of customers had API credentials exposed.
- What to do: Rotate every credential listed in Cursor’s notification and any credential ever sent to Cursor support.
- Tampered builds: Anthropic.claude-code versions 2.1.312 and 2.1.313 were available for about 11 hours on October 29, 2026.
- Status: Fireworks key deactivated, inference moved off Fireworks, credentials rotated, marketplace rebuilt.
In plain English
Cursor, a coding assistant tool, had a security incident in October 2026. An attacker got hold of a key that Cursor had given to Fireworks AI, a company it works with, and that key let the attacker read stored records of coding sessions (called agent traces) that still had passwords and other credentials in them. The attacker used those credentials to break into Cursor’s internal tools and plant malicious code in two versions of a plugin called Claude Code.
What you should do: If you use Cursor and received an email from them, follow the instructions to change your passwords; if you didn’t get an email, Cursor says you were not affected. If you installed Claude Code versions 2.1.312 or 2.1.313 on October 29, follow the cleanup steps in Cursor’s October 29 notice, because simply uninstalling may not be enough.
What happened
Cursor published the root cause of its October 2026 marketplace incident on November 13, 2026, after about two weeks of investigation. The Cursor Team says an attacker obtained an access key that Cursor had issued to Fireworks AI, and that Fireworks exposed that key. Cursor says the last attacker activity it saw was on November 1, 2026, and was blocked.
The key gave read access to a storage bucket Cursor used to share data with Fireworks. That bucket held coding agent traces from sessions outside Privacy Mode, mostly from 2025, and some of those traces still carried unredacted credentials. The attacker used the credentials to reach SaaS tools that Cursor uses, then used a stolen credential to place tampered Claude Code extension builds in the cache serving Cursor’s marketplace.
The tampered builds were Anthropic.claude-code versions 2.1.312 and 2.1.313, which were available for about 11 hours on October 29, 2026. At most 5,000 users requested the cached builds, according to Cursor.
Who is affected
Cursor puts API credential exposure at under 0.01% of customers, and trace exposure at under 0.05% of users who were not in Privacy Mode. Archived non-production source repositories, limited support records and email, product usage metrics, contact details, conversation summaries and credentials users had sent to support were also reachable.
Cursor states that Privacy Mode prompts and code, production systems, customer code repositories and its signed app updates were not affected. Cursor says it contacted affected customers directly, and that anyone who received no notification was not in the exposed traces.
What you should do
Rotate every credential named in Cursor’s notification, which Cursor says it sent to affected admins as lists of specific keys. Rotate anything ever sent to Cursor support, because support records and email were reachable. Credentials that appeared inside coding agent sessions outside Privacy Mode also count, since Cursor says some stored traces kept them unredacted.
Machines that installed Anthropic.claude-code 2.1.312 or 2.1.313 from the Cursor marketplace still need the steps in the October 29 notice, which Cursor warns uninstalling alone may not replace.
What it means
The incident shows that stored coding agent traces can become a security liability when they retain unredacted credentials. Cursor’s November 13, 2026, update names the entry point as a shared Fireworks bucket of stored agent traces rather than Cursor’s production systems, and widens the known data exposure to archived non-production repositories, support records, usage metrics and conversation summaries.
Cursor states the incident was not a vulnerability in Anthropic’s Claude Code product, and that the attacker published the tampered builds using a stolen Cursor credential. Cursor adds that Open VSX never contained the malware, and that its own app and official updates were not signed with exposed keys.
Cursor has deactivated the Fireworks key, moved inference off Fireworks, rotated credentials and rebuilt its marketplace. Cursor directs questions to security@cursor.com.



