Postman Reportedly Installs AI Plugins in Your Coding Agents
In short
- Reports on GitHub say the Postman desktop app installs coding-agent plugins and startup hooks into Claude Code, Codex, Cursor and Copilot without asking.
- Reporters found no evidence of data upload; Postman has not answered in the thread and no fix is announced.
- Check four config files today. The cost is about 10 minutes, not money.
Open the Postman desktop app, and your AI coding agents may get new plugins and new startup instructions. Nobody asked you. A GitHub thread with over 1,400 points on Hacker News says exactly that, and a second user confirmed it after an update on 3 November.
What happened
The issue was opened on 25 October in Postman’s own support repo. The reporter saw that launching Postman desktop 12.33.4 installed a Postman CLI in ~/.local/bin, a Claude Code plugin, a Codex plugin and a startup hook. Other users report a Cursor plugin and a Copilot plugin. One Codex config also got a remote MCP connection to mcp.postman.com, which failed on authentication.
The startup text tells your agent to load a “postman:api-engineer” skill before any non-trivial API task. A second reporter updated from 12.31.3 to 12.35.1 on macOS and logged the same install on 3 November.
Important facts, straight: the reporters say they have no evidence that data was uploaded. The complaint is consent, not theft. And the thread shows no official answer from Postman and no announced fix, as of this morning.
How to use it
If you use Postman with Claude Code, Codex, Cursor or Copilot, check your machine. Look in these places:
- ~/.claude/settings.json and ~/.claude/plugins/installed_plugins.json
- ~/.codex/config.toml
- ~/.copilot/settings.json
- ~/.local/bin/postman
One commenter points to Help, then Postman CLI and Plugins, then Uninstall Postman Coding Agent Plugin. Another says it did not remove the Copilot entry. So verify the files after you click, and edit them by hand if needed. The thread offers no confirmed rollback.
Do the math (my assumption, not a measured number): if the injected instruction adds 3,000 tokens per agent session, 30 sessions a day over 22 working days is about 2 million tokens a month. At $3 per million input tokens that is around $6, roughly €5.50. Small money. The real cost is that someone else writes instructions your agent obeys.
What it means
Agents read config files and obey them, so every tool that can write into those files has power over your agent. Expect more vendors to try this. The rule for your team: treat agent plugin folders like you treat your SSH keys. Check them after every update of any developer tool.
Ivan’s Take
If a tool writes into my agent’s config without asking, it is out until it asks first. I would check those four files today and remove what I did not approve. The data-theft fear is not proven. The consent problem is.
Sources: Postman support repo on GitHub
If it doesn’t save you time or money, you won’t see it here
- Facts, not hypeEvery price and number is checked against the official source
- Always the real priceYou see what it costs before you click
- I say when it’s weakEven when it’s OpenAI or Google
