Discuss brokering credentials for AI agents with Infisical Agent Proxy: proxied services, local and standalone proxies, dynamic secrets for agents, and prompt injection defense.
Use this category for questions, best practices, and discussions around giving AI agents and untrusted code access to APIs without ever handing them real credentials, including:
- Setting up local proxies (agents on your machine) or standalone proxies (agents on your infrastructure)
- Configuring proxied services and mapping hosts to static or dynamic secrets
- Running agents like Claude Code, Codex, and OpenClaw through the proxy via the Infisical CLI
- Credential injection, header replacement, and placeholder keys
- Protecting against credential exfiltration and prompt injection
- Activity logs and monitoring agent traffic
For bugs, use Issues and Bug Reporting; feature ideas go in Feature Requests. For agent access to databases and servers (rather than APIs), see the PAM category.