AegisGate Agent Skill

To set up an LLM API gateway for multi-tenant agent routing, install AegisGate via Docker or local setup, start the service, and register tokens for each upstream provider. You then route client traffic to dedicated v1 token routes or v2 proxy routes.

You can route OpenAI-compatible and Anthropic-compatible endpoints through a single proxy by configuring token routing or direct upstream mode in AegisGate. You then point each client to its dedicated gateway route using the correct baseUrl structure.

To configure client baseUrl for LLM API gateway routing, generate the exact URL structure from AegisGate after registering upstream tokens. You paste this baseUrl into your client configuration, using either v1 token routes or v2 proxy routes, and verify requests work.

Token routing provides better security than direct upstream mode for agent setups by registering specific gateway keys for each tenant. Direct upstream mode is suited for single-user agents, while token routing protects multi-tenant LLM traffic.

After installing an API gateway for LLM traffic, you must complete health checks to confirm gateway readiness. This involves starting the AegisGate service and verifying the service responds correctly before configuring token or direct upstream routing.

You should configure an optional whitelist for your LLM API proxy when you need to restrict access to specific upstream providers or client endpoints. This adds a security layer after token registration using the gateway key.