A developer opened Postman. Then two AI coding agents changed without their permission.

A developer encountered unexpected changes in the AI coding agents Claude Code and OpenAI Codex after opening the Postman desktop application. The developer, referred to as "indicava" on GitHub, discovered that Postman had installed plugins without their permission and attempted to configure its own startup hook. This issue was reported on October 25, following Codex's display of authentication errors and the automatic loading of Postman skills without the developer's consent.

The incident has raised concerns about the ease with which third-party applications can alter AI coding agent behavior without clear user opt-in. Postman's desktop version 12.33.4, when opened on macOS, triggered the installation process, affecting both Claude Code and Codex. The developer found modifications to the coding agents' user-level configuration files, including the enabling of a Postman plugin, a new plugin marketplace, and a trusted session-start hook.

Furthermore, a Postman CLI executable was discovered on the developer's machine, with timestamps coinciding with the installation. Postman's installer reportedly automatically approved a session-start hook on the user's behalf, allowing the app to load the postman:api-engineer skill before handling most API development tasks. However, the developer argues that these changes occurred without their permission, leading to security concerns.

The community has criticized the behavior, calling it "junk" and urging Postman to provide more transparency and user control. The developer seeks Postman to cease installing plugins without permission and to allow for the complete reversal of changes made to agent settings, trusted hooks, MCP connections, and installed CLI tools.