Not clear info on how to properly setup Azure and Azure Cognitive Services Provider

Description

An Issue from Discord where a user could not properly get the Azure provider to work after trying to add it in using the opencode auth login command, due to missing AZURE_RESOURCE_NAME environment variable.

Plugins

None

OpenCode version

1.2.10

Steps to reproduce

Login to Azure Provider using opencode auth login
Put in API Key
Try to use models and get hit with error.

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

This issue might be a duplicate of existing issues. Please check:

How to set up an Azure Foundry (Cognitive Services) model on windows #11792: How to set up an Azure Foundry (Cognitive Services) model on windows — nearly identical issue with the same AZURE_RESOURCE_NAME error when using the Azure Cognitive Services provider after opencode auth login
Azure OpenAI configuration returns "API version not supported" / "Resource not found" despite valid endpoint #12186: Azure OpenAI configuration returns "API version not supported" / "Resource not found" despite valid endpoint — similar Azure provider setup/configuration confusion

If your issue is already addressed in one of these, please consider closing this one. Otherwise, feel free to keep it open with any additional details that differentiate your case.

here's my setup if that helps for Azure Foundry / Azure Cognitive Services:

confirmed working in opencode

Thanks, used that and other tests to improve the docs on cases where the used model is not an openai one

@OpeOginni

For posterity here's the nonFoundry Models (aka /openai/v1 standards) equivalent for anyone interested in future. Not quite sure why one needs resourceName and the other freaks out when resourceName param is used (shrug) > above my intelligence level.

@OpeOginni your context/output limitation setting for Kimi k2.5 in another issue helped so thank you!

What about for those models that don't have OpenAI endpoints routes like Anthrophic's? They seem to not to work for whatever reason:

As for models that have an OpenAI endpoint such as Kimi-K2.5, configuring it as an /openai/v1 model works... until it doesn't and starts exposing the Tool Calls for some reason:

  "provider": {
    "openai-compatible": {
        "options": {
            "baseURL": "https://RESOURCE_NAME.services.ai.azure.com/models",
            "apiKey": "REDACTED",
            "timeout": 300000
          },
        "models": {
            "Kimi-K2.5": {
                "name": "Kimi-K2.5",
                "limit": {
                  "context": 262144,
                  "output": 10000
                },
            }
        }
    }

I personally think that Azure is really not well integrated with OpenCode and must be worked on, or that the documentation is poor at best. My company uses Azure so it's the only thing we have available, and Claude Code (thru Azure Foundry) is costing us hundreds; even Claude has proper documentation on configuring Azure Foundry.

@ktzug

What about for those models that don't have OpenAI endpoints routes like Anthrophic's? They seem to not to work for whatever reason:

@ktzug I just ended up setting two different custom providers: one for OpenAI-style endpoints and another one for Anthropic's. You can use the @ai-sdk/anthropic package:

  "provider": {
    "azure-anthropic": {
      "name": "Azure Foundry Anthropic",
      "npm": "@ai-sdk/anthropic",
      "api": "https://RESOURCE_NAME.services.ai.azure.com/anthropic/v1",
      "models": {
        "claude-sonnet-4-6": {
          "id": "claude-sonnet-4-6",
          "name": "Claude Sonnet 4.6",
          "tool_call": true,
          "attachment": true,
          "reasoning": true,
          "temperature": true
        }
      }
    }
  }

However I gotta agree documentation is severely lacking.

The Azure system is not the cleanest, and is a bit confusing, but I have a PR to improve this flow, and maybe add a special provider for Azure Anthropic, then the Cognitive Service could be for other OpenAI Compatible models

@OpeOginni

Because the support for Azure was always not the best in OpenCode I created a new provider for OpenCode and other products using ai-sdk approach.

https://github.com/ophiosdev/azure-foundry-provider

Highlights

URL-first routing from copied Azure endpoint URLs (no model-name endpoint heuristics)
Supports both chat and responses operation paths
Supports Azure v1 operation paths and /openai/v1 base root with mode-driven routing
Chat transport uses OpenAI-compatible semantics (system role remains system, max_tokens is used)
Request policy control for tools (auto, off, on)
Built-in retries and 429 handling with exponential backoff + jitter
Event-driven waiter queue for maxConcurrent admission (wake-on-release, abort-aware waits)
Optional static quota controls (rpm, tpm, maxConcurrent, maxOutputTokensCap)
Adaptive throttling from Azure x-ratelimit-* headers
Request sanitization for chat history compatibility (removes assistant reasoning_content/reasoning fields)
Automatic chat->responses fallback for model/operation mismatch errors (when mode is not explicitly forced to chat)
Optional observability callbacks (onRetry, onAdaptiveCooldown, onSanitizedRetry, onFallback)
Timeout support via AbortSignal

CC @OpeOginni @rekram1-node

@tmeckel

@tmeckel this seems good, but I was thinking a better solution would be to make providers for each azure openai, azure anthrpoic and azure openai compatible, but lets see what the team thinks of

I'd really like to introduce OpenCode in my agency, but the Azure integration must be done right, otherwise it's a too big of a barrier for users.

@bakaburg1

@bakaburg1 what issue do you have, ik of folks making use of Azure models perfectly on OpenCode. Maybe let me know the issue and what models exactly you are trying to make use of.

Here to help!

To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.

Navigation Menu

Repository navigation

Not clear info on how to properly setup Azure and Azure Cognitive Services Provider #14494

Description

Plugins

OpenCode version

Steps to reproduce

Screenshot and/or share link

Operating System

Terminal

Highlights

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Not clear info on how to properly setup Azure and Azure Cognitive Services Provider #14494

Description

Description

Plugins

OpenCode version

Steps to reproduce

Screenshot and/or share link

Operating System

Terminal

Activity

github-actions commented on Feb 20, 2026

absolutegravitas commented on Feb 21, 2026

OpeOginni commented on Feb 21, 2026

absolutegravitas commented on Feb 22, 2026

ktzug commented on Feb 25, 2026

rifusaki commented on Feb 25, 2026

OpeOginni commented on Feb 25, 2026

tmeckel commented on Mar 6, 2026

Highlights

OpeOginni commented on Mar 6, 2026

bakaburg1 commented on May 28, 2026

OpeOginni commented on May 28, 2026

github-actions commented on Jul 28, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions