Agent decision logs and telemetry data belonging to Warp agentic development environment users were exposed in an attack on a third-party vendor in the project’s infrastructure.
The company’s founder and CEO, Zach Lloyd, estimates that "roughly 1% of Warp users may have been impacted," but external experts are helping with the ongoing investigation.
Warp is a popular developer tool that started as a modern terminal and evolved into an agentic development environment (ADE) that allows developers to direct AI agents to write and test code and deploy software.
The environment provides a terminal, a code editor, task orchestration, and agent review tools in a single workspace. The project's website says the platform has at least 800,000 users.
“On Friday, we became aware of what we believe to be an active and narrow-banded exploitation of a vendor in Warp's infrastructure,” Lloyd stated.
“We detected the operation within hours and were able to remediate and mitigate the impact,” the CEO added.
Although agent traces and telemetry were identified as potentially affected data, the company did not clarify whether the exposed information included source code, prompts, credentials, or other sensitive details.
Analysis of the incident suggests that the attack was powered by "an agent swarm" that relied on "a very sophisticated penetration and escalation technique."
Lloyd said that Warp has contacted the third-party vendor about the vulnerability, but no details have been shared. The platform has notified users identified to be impacted.
The cybersecurity firm CrowdStrike has also been engaged to assist with the investigation, while the platform has also informed law enforcement authorities.
The incident at Warp follows an incident at AI coding platform OpenCode, which notified users last week that a security breach exposed stored API keys used with external providers.
The incident affected a small percentage of OpenCode users, about 1.4%, and led to unauthorized charges of up to $1,700 for some.
Warp users are advised to remain vigilant, monitor account activity, and treat unsolicited communications with increased caution.
Once attackers have valid credentials, only 37% of their actions are blocked
Overall prevention scores can hide what happens after initial access. Once attackers are using valid credentials, prevention drops sharply.
The Blue Report 2026 measures defenses technique by technique across 338 million simulations run in customer production environments.
Get the report