Postman Quietly Installs Plugins into Claude Code and Codex, Then Goes Silent for 8 Days
The incident exposes how fragile the trust model is for AI coding assistants: a single configuration file can be silently rewritten by any installed tool. For developers who rely on Postman alongside Claude Code or Codex, it means auditing configs, questioning vendor intent, and reconsidering which tools get such privileges.
A developer using Postman 12.33.4 discovered on October 24 that the tool had silently installed its plugins into Claude Code and Codex. For logged-in existing users, a single startup triggers the install: it adds a plugin marketplace and a trusted startup instruction to Codex config, enables Postman's plugin in Claude Code settings, and creates a postman command with a remote connection in the home directory. The startup instruction tells Claude Code to load Postman's API specification for interface-related tasks.
The issue report, filed on October 26 with the word "severe" in the title, went without official response for eight days until comments poured in on November 3 and a Hacker News thread (1,400+ points, 460+ comments) spread it on November 4. The Chinese developer community followed suit with a summary post that reached hundreds of thousands of views. Community reactions split: many are leaving for tools like Bruno, Insomnia, and Milkman; others argue the real problem is the trust model where a single config line gives any software control over AI assistants; a few point out that the widely shared claim about API data sharing has no evidence.
Verified facts include the auto-install for logged-in users, config changes, and the postman command. Whether the startup instruction actually ran is contested between two individuals, and no log supports data upload. Users can check with three grep commands and uninstall via the Postman client menu under Help > "Postman CLI and Plugins", though some report Copilot removal is incomplete. This incident follows Postman's earlier backlash over cutting free team plans and a June 2024 security breach where customer lists were taken from its sales system.
The incident reveals that AI coding assistants trust configuration files as absolute authority; any software that can write a line into ~/.claude or ~/.codex gains silent control over tool behavior.
Postman's history of user-hostile moves, from pricing cuts to a real security breach, makes this pattern especially damaging to its credibility among developers already looking for alternatives.
The community's three-way split mirrors a broader tension: some want practical alternatives, others see the systemic trust flaw, and a few correctly refuse to inflate unverified claims.
The gap between confirmed config changes and unconfirmed data sharing shows how quickly speculation can outrun evidence in a viral incident.