Agent Vault lets MCP agents authenticate without seeing credentials
This pattern reduces the main risk of giving coding agents browser or API access: leaked secrets in prompts, logs or model outputs. Teams running Codex or other MCP-compatible agents can adopt it today with any vault that exposes an MCP server. It is a credential-handling design, so it works regardless of which model sits behind the agent.
A workflow posted on 5 Jan 2026 explains how to connect Codex to a credential vault through MCP so agents can log in to websites without ever seeing passwords. The agent is given placeholder credentials instead of real ones. When Codex attempts to authenticate, Agent Vault's proxy swaps the placeholder for the real credentials at the point of use. The model completes the login normally while the password stays hidden from it.
This will let your agents use credentials without sharing them with the model.
For example, you can connect Codex to the vault using MCP and give it access to placeholder credentials to log in to a website.
When Codex tries to log in, Agent Vault's proxy will replace the placeholder with the real credentials. Codex will never see the password, and the agent can authenticate normally.
