Feature Request — Support for Claude Models on Azure AI Foundry

[thefunny1983]

Hi LibreChat team,

I’d like to request support for Claude models deployed in Azure AI Foundry in Librechat.

Background
Currently, I’ve tried setting up LibreChat with Claude hosted on Azure AI Foundry using the Anthropic API endpoint format provided by Azure. I tried multiple things, such as removing the defaultParamsEndpoint, playing with headers and such, but could not get to a working situation.

Example config (redacted for sensitive info):

custom:
- name: "Claude"
  baseURL: "https://<resource-name>.services.ai.azure.com/anthropic/v1"
  apiKey: <redacted>
  iconURL: "anthropic"
  models:
    default: ["claude-sonnet-4-5"]
    fetch: false
  titleConvo: true
  titleModel: "claude-sonnet-4-5" (I know this is not very friendly for the environment ;-), but this is just for test) 
  summarize: false
  summaryModel: "claude-sonnet-4-5"
  forcePrompt: false
  modelDisplayLabel: "Claude Sonnet 4.5"
  customParams:
    defaultParamsEndpoint: "anthropic"
  headers:
    x-api-key: <redacted> 
    anthropic-version: "2023-06-01"

Issue

Requests fail with the following error:

Something went wrong. Here's the specific error message we encountered:
404 Requested API is currently not supported
Troubleshooting URL: https://js.langchain.com/docs/troubleshooting/errors/MODEL_NOT_FOUND/

It appears LibreChat currently does not support Azure’s Anthropic API routing or model name mapping.

Request
Add first-class support for Claude models in Azure AI Foundry. Update documentation with example configs for Azure AI Foundry Claude deployments (similar to OpenAI Azure integration).

It was released one day ago: Announcement
And here: https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry

[danny-avila]

It is possible to use Anthropic models via Azure foundry already as a full replacement of the Anthropic endpoint (no custom endpoints, no azure config).

This works for me:

ANTHROPIC_REVERSE_PROXY=https://your_endpoint.services.ai.azure.com/anthropic/
ANTHROPIC_API_KEY=your_key
ANTHROPIC_MODELS=claude-haiku-4-5,claude-sonnet-4-5,claude-opus-4-1

At this point, there is enough customization for custom endpoints that it should be possible there too (by specifying what the underlying provider for API interactions to be used), but it will require a bit more work and a specific configuration afterwards.

[dycolin]

Working well, thanks

[thefunny1983]

Great Danny, works perfectly. Thanks!

[kkrkarthik]

@danny-avila Please add this step to the documentation for anyone who needs it.

[danny-avila]

As mentioned in my earlier comment, this will also be possible by way of this feature: #10655

[RepLicanT-UHD]

It's absulutelly insane given that there are 450 000 TPM limits there for all Claude family. Works fine, thanks a lot!