LLM Gateway
Direct gateway deprecation and migration to LLM integrations
Direct LLM Gateway access through the VM metadata endpoint is deprecated and
is unavailable to accounts created on or after 2026-08-03 00:00 UTC. Use the
default [LLM integration](/docs/integrations-llm) at
`https://llm.int.exe.xyz` instead.
The `exe.dev LLM gateway` provider source within an LLM integration remains
supported. It provides exe.dev-managed access to Anthropic, OpenAI, and
Fireworks models. Your subscription includes a monthly token allocation, and
you can purchase additional tokens at [https://exe.dev/user/shelley](https://exe.dev/user/shelley).
See the [full list of supported models](/llm-gateway-models) ([JSON](/llm-gateway-models.json)).
New accounts get a default [LLM integration](/docs/integrations-llm) named
`llm`, attached to `auto:all`. That integration exposes managed models at
`https://llm.int.exe.xyz` inside attached VMs, with no provider API keys stored
on the VM.
The gateway is only for exe.dev-managed model access. If you want to use your
own provider API key or a ChatGPT subscription, configure those as provider
sources on an [LLM integration](/docs/integrations-llm); those sources are not
part of the gateway or its token allocation.
Use the [LLM Integration guide](/docs/integrations-llm) to configure provider
sources, connect a ChatGPT subscription, attach the integration to VMs, or use
the integration with Shelley, Codex, Claude Code, and curl.
## Shelley
[Shelley](/docs/shelley/intro) automatically discovers attached LLM
integrations through the `reflection` integration. On new accounts, Shelley
sees the default `llm` integration and shows its managed models in the
`Model:` picker without custom model setup.
If Shelley does not show the integration models, see
[Use with Shelley](/docs/integrations-llm#use-with-shelley).
## Deprecated direct endpoint
Do not use the legacy metadata endpoint:
```
http://169.254.169.254/gateway/llm/<provider>
```
Use the default LLM integration instead:
```
$ curl https://llm.int.exe.xyz/v1/models
```
See the [LLM Integration guide](/docs/integrations-llm#use-from-a-vm) for
provider-specific examples.