Beta: The LLM Gateway is in beta. APIs and features may change as we iterate.
moonshotai/kimi-k3 uses Gateway Credits. A model ID that starts with a configured bring-your-own-key provider, such as anthropic/claude-opus-5, uses that provider’s secret instead.
Base URL
Point any supported client at the standard gateway API:Authenticate with your LangSmith API key as a bearer token.For regional base URLs, see Regional gateways.
Available models
Select a hosted model by ID in the request body. Model IDs are case-insensitive.
List every model available to your workspace, including configured bring-your-own-key providers and hosted models, with the standard model-list endpoint:
model field of a prompt request.
Prerequisites
Before using Gateway Credits:- Your organization is on a paid plan (Developer, Plus, Startup, or Premier).
- You have a workspace-scoped LangSmith API key attached to a role with
gateway:invokeandworkspaces:readpermissions. See Admin setup if you are unsure.
Make a call
Point an OpenAI-compatible client athttps://gateway.smith.langchain.com/v1, authenticate with your LangSmith API key, and set model to a hosted model ID.
Supported endpoints
Hosted models use the same standard API formats as bring-your-own-key models:
For request examples and translation behavior, see API formats.
Pricing
Gateway Credits are available on all paid plans besides Enterprise. See the pricing page for plan details and current rates. Gateway Credits are denominated in LangChain Credit Units (LCUs) at $1.50 per LCU; each call consumes LCUs based on token usage. Standard gateway spend policies apply to hosted-model traffic, so any organization, workspace, API key, or user cap you have configured also governs Gateway Credit usage. You can control Gateway Credit consumption with the same tools you use for bring-your-own-key providers. For example, cap a specific API key at $200/month across every provider, or set a workspace-wide daily limit that includes hosted-model calls.Tracing
Like all gateway traffic, hosted-model calls are traced to LangSmith. For where traces land and how to control access to them, see Traces, Engine, and access control.Next steps
- Quickstart: make your first gateway-proxied call.
- API formats: call models through Chat Completions, Messages, or Responses.
- Spend policies: add cost limits to Gateway Credit usage.
- Direct model access: use provider-native APIs and model IDs.
Connect these docs to Claude, VSCode, and more via MCP for real-time answers.

