Gateway Adapter

The FORG Gateway captures AI usage at the proxy layer. Instead of installing anything on each client, you route LLM requests through the gateway, which creates synthetic sessions with model, token, and cost metadata. It works with any provider or OpenAI-compatible endpoint with zero client-side install. There is noforg connect gateway command and no MCP configuration for this adapter.

How it works

Point your AI clients' base URL at the FORG Gateway endpoint. The gateway forwards each request to the upstream provider, records metadata about the exchange, and returns the response unchanged. Requests are attributed to synthetic sessions so they appear in your dashboard alongside adapter-based sessions.

Install

Route LLM traffic through the FORG Gateway by setting the gateway as your API base URL (for example, via your SDK's base_url/OPENAI_BASE_URLconfiguration). Gateway virtual keys and hard server-side caps are available on the Team and Enterprise plans — see the Gateway overview and supported providers.

What gets captured

• Synthetic sessions for every request routed through the gateway
• Model, token, and cost metadata
• Provider-agnostic coverage (any OpenAI-compatible endpoint)

Notes

• The gateway is a proxy, not an MCP server — no per-client install is required.
• Metadata-only: request and response bodies are forwarded, not stored.