If you run Postman on your machine alongside Claude Code or OpenAI Codex, you may want to audit what those tools currently have loaded.
A developer reports that Postman’s desktop app installed plugins into both Claude Code and OpenAI Codex without asking. No prompt. No confirmation dialog. The installations happened when they opened the app.
What Was Installed
The app pushed its own plugins into two separate AI coding agents. It also attempted to approve its own startup hook inside OpenAI Codex. That startup hook tells Codex to use Postman’s tools for API development tasks, meaning Postman was trying to insert itself into the Codex workflow automatically, without the user initiating it.
Why This Matters
AI coding agents like Claude Code and Codex operate with elevated access to your codebase, your terminal, and in some configurations your filesystem. Any plugin that gets silently dropped into those environments has the potential to influence what the agent does, what context it receives, and what tools it calls.
This isn’t a theoretical concern. A startup hook that redirects an AI agent toward a specific vendor’s toolset changes how that agent behaves, and the developer in this case didn’t choose that change.
What to Check
If you have Postman installed on macOS or any other platform alongside Claude Code or Codex, it’s worth checking your MCP server configurations and agent plugin lists. Look for any Postman-related entries you didn’t add manually.
- Claude Code: check your
.claude/settings.jsonor MCP server config for unexpected entries - OpenAI Codex: review any startup hooks or registered tools for Postman references
The developer’s report was covered by The New Stack. Postman has not publicly responded to the specific claims as of the time of publication.
