Хочете будувати кар’єру в Data Science та AI? Vodafone Big Data Lab відкриває новий набір —>
Наталя ХандусенкоAI Eng
4 November 2026, 12:21
2026-11-04
Postman Caught Secretly Installing AI Plugins: Desktop App Altered Claude Code and Codex Without Developer Consent
Developers have raised questions about the security and control of artificial intelligence after the popular tool Postman was found to be secretly installing plugins into AI assistant systems.
The incident became public after a developer under the nickname indicava published a report on GitHub. On October 25, he noticed that the Codex AI agent began to issue authentication errors and constantly load Postman tools without a corresponding request. As it turned out, after opening the desktop version of Postman 12.33.4 on macOS, the program spontaneously integrated into the Claude Code and OpenAI Codex environments.
Developers have raised questions about the security and control of artificial intelligence after the popular tool Postman was found to be secretly installing plugins into AI assistant systems.
The incident became public after a developer under the nickname indicava published a report on GitHub. On October 25, he noticed that the Codex AI agent began to issue authentication errors and constantly load Postman tools without a corresponding request. As it turned out, after opening the desktop version of Postman 12.33.4 on macOS, the program spontaneously integrated into the Claude Code and OpenAI Codex environments.
What exactly did Postman do?
The application logs for October 24 recorded an automatic update that added the Postman extension to Claude Code and Codex without the user's knowledge, The New Stack reports . At the same time, the program arbitrarily modified the configuration file ~/.codex/config.toml, enabling its own plugin, adding a new marketplace, and installing additional CLI tools.
Moreover, Postman bypassed the standard security check: instead of receiving manual confirmation from the developer, the installer generated trust hashes itself and automatically approved its own autorun script in Codex.
Finally, the settings specified a connection to a remote MCP server ([https://mcp.postman.com/mcp](https://mcp.postman.com/mcp)), which, due to the lack of authentication, caused systematic errors in the operation of the AI assistant.
Thanks to these manipulations, Postman forced the Codex AI agent to use its own tools by default to perform most API development tasks, ignoring the developer's initial settings.
Reaction from the community and security experts
This behavior caused a wave of outrage among IT professionals at GitHub and X.
Cybersecurity researcher Frank Denis called such integrations "garbage" and criticized the app's attempts to grant itself trust rights.
Other developers emphasize that such actions are unacceptable and pose serious risks to corporate security.
There were also speculations online about the possible interception and sending of Postman API data (in particular from developer Siddharth Garimella), but the report on GitHub found no evidence of a data leak - it only refers to an unauthorized change in configurations.
What developers require
The report author demands that Postman stop forcing plugin installation and return control over script approval to users. Additionally, the standard uninstallation of Postman currently does not completely clean up modified AI agent configurations, saved MCP connections, and installed CLI tools.
Experts recommend that developers immediately check their Claude Code and Codex settings for unfamiliar plugins, autorun hooks, or third-party MCP connections.
Additional concern was also raised by an experiment by the startup Cathedral, whose AI agents took less than 12 hours to penetrate Anthropic's internal systems through vulnerabilities in corporate software and gain access to Slack correspondence, meeting records, and employee salary data.
Як нейромережі бачать вільну та незалежну Україну? Тест dev.ua
Нейронні мережі для генерації зображень бачать світ по-своєму, їхню логіку зрозуміти часом зовсім неможливо. Але таки хочеться. На честь Дня Незалежності України редакція dev.ua вирішила провести невеликий експеримент.
Ми задали чотирьом різним нейронним мережам п’ять однакових запитів: «прапор України», «День Незалежності України», «український Крим», «перемога України» та «українці». Отриманими результатами ми ділимося з вами нижче.
У TikTok тепер можна генерувати фон за допомогою нейромережі. Ми протестували її та ділимося результатами
У TikTok з’явилася нова функція «Розумний фон». З її допомогою як фон для тіктоків можна підставляти згенеровані нейромережею зображення. Редакція dev.ua протестувала цю технологію і ділиться своїми враженнями.