New partnership between 1Password and OpenAI provides access tools, privacy

A new partnership could change how AI agents develop code by instituting guardrails on their behavior.

1Password is working with OpenAI to provide an access layer for the latter company’s Codex coding platform. By connecting Codex to 1Password’s existing environments, the partnership is solving a credential problem that has spiraled out of control, 1Password CTO Nancy Wang told IT Brew.

“AI coding agents are creating a new identity surface—which means that you now have coding agents that are acting on behalf of your developers, doing things that your human developers wouldn’t normally do,” Wang said.

Details. Codex will run prompted tasks through a series of 1Password permissions. At no point will Codex configure an environment without getting explicit user permissions at every step of the way. The platform also ensures that your identity access is safely protected, as Wang explained.

“We’re not just giving credentials to the agent, so it can have it forever; we’re injecting it into our local MCP server, but not injecting it in a plain text way,” Wang said. “It’s via our dev environments product, which is a mounted in-memory file, so nothing is stored on disk.”

Not the first to notice. Identity access management for AI systems has been top of mind for security-focused IT pros for some time now. Cisco SVP and GM of Infrastructure and Security Tom Gillis told IT Brew in summer 2025 that having insight into processes on a user’s machine can help differentiate between humans and agents. This April, DTEX CEO Marshall Heilman warned us about the potential for identity access to spin out of control.

“In this new world we find ourselves in, when you start having AI agents that can act autonomously, they are acting at machine speed, at the speed that computers operate,” Heilman told IT Brew. “Humans having to go and look through a bunch of alerts and trying to figure out what happened and then responding—you’re just so far behind what happened that if there is a malicious action or there’s something negligent that happens, you’re too far behind it to really prevent damage.”

Locking in. Leveraging 1Password’s workflow oversight prevents company secrets and sensitive files from bleeding into OpenAI’s servers or other accessible networks. Agents are too unpredictable to allow them to determine their own access and trust they’ll behave as directed, Wang said.

“Agents don’t always follow instructions; in fact, there are many examples out in the wild of agents ignoring [what they’re told],” Wang said.

For organizations using Codex, 1Password and OpenAI say this new partnership will free up time to get tasks done and streamline workflows. Added efficiency from turning management over to 1Password’s system should also give security staff more bandwidth for keeping data safe.

Ultimately, you can’t fully trust an autonomous system, no matter how much it manages its behavior effectively—Wang told IT Brew that’s the driving force behind this initiative.

“This goes back to accountability and responsibility, and this is one of the big reasons why agents need their own identity, because it’s not always built off of the human,” Wang said.