OPEN LINK ↗
// 1h agoSECURITY INCIDENT
Postman Auto-Installs AI Plugins, Trusts Hook
A GitHub issue alleges Postman Desktop 12.33.4 silently installed Codex and Claude plugins, a CLI, remote MCP configuration, and a broad startup hook on macOS. The report documents unwanted configuration changes and attempted hook self-approval, but does not establish API-data exfiltration.
// ANALYSIS
This is a serious consent and trust-boundary failure, even without proof that data was shared. Postman’s agent integrations can access workspaces, collections, specifications, mocks, and monitors, making silent installation especially consequential.
- –The reported installer modified global Codex and Claude Code configurations outside Postman.
- –The startup hook broadly routed API-engineering tasks through Postman skills.
- –Installer-side hook approval undermines the agent host’s user-controlled trust model.
- –Postman’s MCP integration can access substantial API-development context once authenticated.
- –Explicit opt-in, transparent scope, and complete rollback should be mandatory for integrations of this kind.
// TAGS
postmansecurityagentmcpapidevtool
DISCOVERED
1h ago
2026-11-03
PUBLISHED
5h ago
2026-11-03
RELEVANCE
8/ 10
AUTHOR
sidgarimella