LLM Gateway Improvements: November 16, 2026

Shared key ownership

Previously, management of LLM Gateway keys was restricted to only Account Owners and a single key owner — the user who created the key. Now, you can set multiple key owners, each of whom:

• Can modify key spend limits and receive spend alert emails

• Can view request logs and metadata

• Can update or revoke the key

To learn more, see: LLM Keys

Decisions API

The LLM Gateway now supports a Jev- and Laya-compatible Decisions API (POST /v1/decisions), backed by chat models already available via the Gateway. To learn more about how to use the Decisions API, see Supported APIs. To see which models you can use with the Decisions API, refer to the Supported Models table.

Full support for Claude, Codex, and other agent harnesses

Previously, when using LLM Gateway with Claude Desktop, Claude Code and other agent harnesses, it was necessary to configure an explicit list of models (e.g. by setting the inferenceModels variable in Claude Desktop's configuration). Now, LLM Gateway supports turnkey integration with Claude Desktop, Claude Code, Codex, Pi, Grok Build and more. Just point your harness at LLM Gateway, and all compatible models are discovered automatically and ready to use from the harness. To learn more, see Agent Harness Support.