OpenAI added a beta Developer mode to ChatGPT. This feature allows full read and write support of MCP (Model Context Protocol), though the documentation describes it as dangerous. Edwin Arbus, developer community lead at OpenAI, said “in developer mode, developers can create connectors and use them in chat for write actions (not just search/fetch). Update Jira tickets, trigger Zapier workflows or combine connectors for complex automations.” that the initial beta is limited in that the developer mode cannot be used in team workspaces or project chats.
Arbus showed how the feature can link ChatGPT with Stripe so that AI can generate invoices and send in response to a prompt. There is a confirmation button that appears before an action, but it can be disabled by a trusted MCP Server by checking the option to “remember for this conversation.”
In principle, users are enthusiastic. said one. Other comments on X say that the feature will make ChatGPT “10x more practically useful” and that it “changes the game for production workflows.”Some users have experienced bugs, either with connectors that don’t work, or return fictional information. One user connected ChatGPT with Atlassian’s Jira tool and asked it to summarize a problem. He reported the tool was called, returned information, but that “GPT completely make up the summary, it’s something completely different than in the Jira issue.”
more serious problems are possible. The documentation says that the feature is “powerful but dangerous, and is intended for developers who understand how to safely configure and test connectors. When using developer mode, watch for prompt injections and other risks, model mistakes on write actions that could destroy data, and malicious MPCs that attempt to steal information.”
ChatGPT Developer mode comes with a warning.
ChatGPT’s popularity among mainstream users makes it more likely for risks to be not understood. Simon Willison, Django cofounder and AI enthusiast, said “Wow this is dangerous,” . Google and Zed push protocols to free AI agents from VS Code’s grasp
In a comment on X, it was described that the feature could be used to exfiltrate emails data, sending the victim a Calendar This could be triggered if the user asks ChatGPT to “help prepare for their day by looking at their calendar.”
The industry adoption of MCP continues at a rapid pace, and vendors that do not support it may find themselves left behind. Hacker News user “Calling out ChatGPT specifically here feels a bit unfair,” noted that other vendors had already shipped similar features. (r)
