1Password & Keycard Secure AI Agent Credentials

1Password and Keycard are introducing new tools to secure how AI coding agents handle credentials. This comes as AI agents are increasingly writing and deploying code autonomously, requiring access to real systems. 1Password has expanded its collaboration with OpenAI. Developers can now grant OpenAI's Codex coding AI agent access to credentials directly within their coding workflows. This keeps sensitive information out of prompts and code. 1Password uses a new MCP Server to inject secrets at runtime, ensuring they are not written to disk and are only available for the duration of an execution or session. Nancy Wang, CTO of 1Password, emphasizes that just-in-time credentials are the only viable security model for AI-native development. OpenAI's Nick Steele notes this integration simplifies agentic development, allowing teams to ship faster while protecting sensitive credentials. Keycard, another identity and access management provider, has also launched a new feature called Keycard for Multi-Agent Apps. This aims to extend its platform for multi-agent applications. These developments are important because they address a growing security problem as AI agents become more prevalent in software development.