A recent report has revealed that Postman, a popular API client, silently auto-installs AI plugins and attempts to share all API data without user consent. This has sparked concerns over trust and data privacy, as users are unaware of the changes being made to their coding agents and API workflows. The issue was discovered on a Mac, where the user found that Postman had installed a Claude Code plugin and a Codex plugin, and had also enabled a remote MCP connection.
What happened
The user reported that upon opening Postman, they noticed unexpected changes to their coding agents, including recurring MCP authentication warnings and Codex repeatedly routing API work through Postman skills. An investigation found that Postman had auto-installed a user-wide plugin, a startup instruction hook, and a remote MCP connection, all without the user's knowledge or consent. The plugin installation logs show that Postman's own log explicitly records the automatic installation of the plugins.
Technical analysis
The technical analysis of the issue reveals that Postman's auto-installation of AI plugins is a result of a script that runs when the user opens the application. The script installs the plugins and enables the remote MCP connection, all without prompting the user for consent. The plugin code shows that the hook registers a SessionStart command with a matcher that prints a message telling the user to load the Postman skill and follow it. The installer code also attempts to approve its own hook installer, which raises concerns over trust and data privacy.
Why it matters
The issue matters because it raises concerns over trust and data privacy. Users expect to have control over their coding agents and API workflows, and the silent auto-installation of AI plugins without consent undermines that trust. The potential consequences of the auto-installation are unclear, but it could lead to unintended changes to the user's coding agents and API workflows, potentially compromising data security.
The SoftCodeX view
We expect that this issue will prompt a re-evaluation of Postman's installation and update processes, with a focus on transparency and user consent. In our view, the silent auto-installation of AI plugins without consent is a significant concern that needs to be addressed to maintain user trust.
What to do now
Engineering and business teams should take immediate action to review their Postman installations and update processes. Users should check their Postman logs and configurations to ensure that they have not been affected by the auto-installation of AI plugins. Additionally, teams should consider implementing measures to prevent similar issues in the future, such as requiring explicit user consent for plugin installations and updates. By taking these steps, teams can help maintain user trust and ensure the security and integrity of their coding agents and API workflows.
Frequently asked questions
What is the issue with Postman's AI plugin installation?
The issue is that Postman silently installs AI plugins without user consent, altering coding agents and attempting to share API data.
How does the auto-installation affect users?
The auto-installation changes the user's coding agents, routes API work through Postman skills, and attempts to approve its own instruction hook, raising concerns over trust and data privacy.
Is data exfiltration confirmed?
No, the report does not claim to establish data exfiltration, but rather unwanted integration and instruction changes.
Loading comments…