Postman silently auto‑installs AI plugins and tries to share all API data
agents claude startup
| Source: HN | Original article
Postman has been found to silently auto‑install AI plugins that attempt to share all API data.
Postman has come under scrutiny after users discovered that the API development platform silently installs AI‑driven plugins and attempts to transmit the contents of their API projects. A Mac user reported that simply launching Postman triggered a cascade of unexpected behaviours: repeated MCP authentication warnings, Codex routing API calls through Postman‑generated “skills”, and the appearance of a Claude Code plugin that had not been explicitly enabled. Further investigation revealed a user‑wide Postman plugin, a startup instruction hook, a remote MCP connection, and a separately installed Postman CLI—all added without clear user consent.
The incident highlights a growing tension between the convenience of integrated AI assistants and the security of development environments. Postman’s own documentation confirms that its Enterprise package supports silent, system‑wide installations that can be configured by administrators, a feature that can be misused to push additional components onto workstations. The newly announced Postman Plugin for Claude Code, which links a workspace to an AI coding agent, exemplifies how tightly AI tools can be woven into the API lifecycle, but also how they can become vectors for data exfiltration if not properly guarded.
Security experts warn that autonomous agents capable of accessing API specifications, environments and test results could inadvertently expose sensitive business logic, credentials or proprietary data. The episode underscores the need for tighter controls around AI plugin distribution, clearer consent mechanisms, and robust monitoring of outbound connections from development tools.
Postman has not yet issued a formal statement, but the company’s “AI Engineer” and “Context Graph” features are marketed as safeguards for autonomous agents. Watch for an official response, potential patches to the silent‑install pathway, and broader industry discussion on how API tooling should balance AI functionality with data protection.
Sources
Back to AIPULSEN