Overview
- A developer reported that opening Postman desktop on October 24, 2026 triggered writes to Codex and Claude Code user configs that enabled a Postman plugin, added a marketplace entry, and set a trusted session-start hook.
- The report shows Postman’s installer can add a trusted hash to Codex’s configuration using a config/batchWrite call so the hook is treated as approved without the agent showing a manual allow prompt.
- Files and timestamps on the machine also show a Postman CLI and a configured remote endpoint at https://mcp.postman.com/mcp that repeatedly failed to authenticate but no evidence was presented of API data being exfiltrated.
- Security researchers and developers have criticized the behavior and asked Postman to stop automatic installs for existing users and provide a reliable way to fully undo changes; Postman had not responded to inquiries by publication.
- For now developers are advised to check agent config files such as ~/.codex/config.toml and ~/.claude for unfamiliar plugins, hooks, or MCP entries because current uninstall steps may not remove trusted hashes or hook settings.