Skip to main content
AI tools and agents increasingly need credentials like API keys, tokens, and passwords to complete tasks. When you supply these secrets in plaintext, they can leak into LLM context, source control, or config files. 1Password gives you secure methods to use secrets in agentic workflows without exposing them in prompts, config files, or agent memory.

Workflow: Secure AI access

Learn how to get started using 1Password to secure your agentic workflows.

Validate your Environments setup with hooks

Validate locally mounted .env files from 1Password Environments before supported AI coding agents execute shell commands.

Use the 1Password plugin for Claude Code

Install one plugin to add .env file validation, the Environments agent skill, and MCP server configuration to Claude Code.

Use the 1Password plugin for Cursor

Install one plugin to add .env file validation, the Environments agent skill, and MCP server configuration to Cursor.

Secure agent access with NVIDIA OpenShell

Give sandboxed coding agents scoped, time-limited access to secrets. The agent works with placeholders and never sees secret values.

Manage Environments from your MCP client

Use the 1Password Environments MCP Server to create and manage Environments with secure authorization prompts, without returning secrets to the agent.

Secure MCP server config files

Replace plaintext tokens in your mcp.json file with environment variables loaded from 1Password at runtime.

Authenticate AI command-line tools with biometrics

Use shell plugins to authenticate AI CLIs like OpenAI CLI or Claude Code with your fingerprint instead of storing API keys in shell profiles.