feat: add Azure AI Foundry provider support by alexpwrd · Pull Request #6969 · openclaw/openclaw

alexpwrd · 2026-02-02T07:36:20Z

Summary

Adds native Azure AI Foundry (now Microsoft Foundry) provider support, addressing [Feature Request] Add native Azure OpenAI / Azure AI Foundry as model provider #6056
Two provider IDs sharing one API key: azure-ai-foundry (OpenAI-compatible) and azure-ai-foundry-anthropic (Anthropic-compatible)
Static model catalog with standard Azure model IDs (GPT-5/4.1/4o series, Claude 4.5/4/3.5 series, o3/o4 reasoning models)
Auth fallback chain: AZURE_AI_FOUNDRY_API_KEY → AZURE_FOUNDRY_API_KEY → AZURE_OPENAI_API_KEY
Endpoint auto-discovery via AZURE_FOUNDRY_OPENAI_ENDPOINT / AZURE_FOUNDRY_ANTHROPIC_ENDPOINT
Custom deployment names supported via explicit models.providers config
Documentation and test included

Notes

Microsoft renamed Azure AI Foundry to "Microsoft Foundry" in Jan 2026. Provider ID uses azure-ai-foundry since the API endpoints still use Azure domains. Can be aliased/renamed later.
Model costs default to 0 (typical for Azure credit/sponsorship users). Users on pay-as-you-go can override via explicit config.
Follows the static catalog pattern used by Minimax/Moonshot providers.

Test plan

Type-check passes (tsc --noEmit)
Unit test passes (vitest run)
Tested locally with Azure AI Foundry endpoint (Claude Opus 4.5 via Anthropic-compatible API)
Verified openclaw models list shows all catalog models with auth
Verified gateway WebChat works end-to-end with Azure-hosted model

Closes #6056

Greptile Overview

Greptile Summary

This PR adds a new Azure AI Foundry provider integration by:

Extending auth resolution to support azure-ai-foundry / azure-ai-foundry-anthropic with an env-var fallback chain.
Auto-registering two implicit providers (OpenAI-compatible and Anthropic-compatible) based on AZURE_FOUNDRY_*_ENDPOINT env vars and a static model catalog.
Adding provider documentation and a small unit test.

The new providers plug into the existing resolveImplicitProviders flow (used by ensureOpenClawModelsJson to write/merge models.json) and into the shared env auth resolution in src/agents/model-auth.ts.

Confidence Score: 2/5

Not safe to merge as-is due to incorrect API key resolution for implicit providers.
The Azure AI Foundry provider addition is straightforward, but resolveImplicitProviders currently assigns the env var name (e.g. AZURE_AI_FOUNDRY_API_KEY) as the apiKey value, which will break authentication when the implicit providers are used. Aside from that, the changes are localized and consistent with existing provider patterns.
src/agents/models-config.providers.ts (implicit provider apiKey resolution), plus a quick scan of other implicit providers using resolveEnvApiKeyVarName.

_{(2/5) Greptile learns from your feedback when you react with thumbs up/down!}

Adds native support for Azure AI Foundry (now Microsoft Foundry), allowing users to connect OpenAI and Anthropic models hosted on Azure under a single API key. - Two provider IDs: azure-ai-foundry (OpenAI) and azure-ai-foundry-anthropic (Anthropic) - Static model catalog with common Azure model IDs (GPT-5/4.1/4o series, Claude 4.5/4/3.5 series, o3/o4 reasoning) - Auth fallback chain: AZURE_AI_FOUNDRY_API_KEY → AZURE_FOUNDRY_API_KEY → AZURE_OPENAI_API_KEY - Endpoint discovery via AZURE_FOUNDRY_OPENAI_ENDPOINT and AZURE_FOUNDRY_ANTHROPIC_ENDPOINT - Custom deployment names supported via explicit models.providers config Closes openclaw#6056

greptile-apps

_{2 files reviewed, 2 comments}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-02-02T07:38:44Z

src/agents/models-config.providers.ts

+  const azureFoundryKey =
+    resolveEnvApiKeyVarName("azure-ai-foundry") ??
+    resolveApiKeyFromProfiles({ provider: "azure-ai-foundry", store: authStore });
+  if (azureFoundryKey) {


azureFoundryKey is being set to the env var name (via resolveEnvApiKeyVarName) rather than a resolved credential value. That means the provider config may contain the string name of an env var instead of an actual token, leading to authentication failures when the implicit provider is used.

This should likely mirror the other implicit providers here by resolving a concrete key value for apiKey, and only using resolveEnvApiKeyVarName when you intentionally want an env-var placeholder.

Also appears at src/agents/models-config.providers.ts:632-634, 647-649, 654-656, 661-663, 676-678, 684-686.

Prompt To Fix With AI

This is a comment left during a code review. Path: src/agents/models-config.providers.ts Line: 692:695 Comment: `azureFoundryKey` is being set to the env var *name* (via `resolveEnvApiKeyVarName`) rather than a resolved credential value. That means the provider config may contain the string name of an env var instead of an actual token, leading to authentication failures when the implicit provider is used. This should likely mirror the other implicit providers here by resolving a concrete key value for `apiKey`, and only using `resolveEnvApiKeyVarName` when you intentionally want an env-var placeholder. Also appears at src/agents/models-config.providers.ts:632-634, 647-649, 654-656, 661-663, 676-678, 684-686. How can I resolve this? If you propose a fix, please make it concise.

This follows the same pattern as all other implicit providers in this file (Minimax, Moonshot, Ollama, Venice, Xiaomi, Synthetic) — see lines 633, 648, 661, 676, 684. resolveEnvApiKeyVarName returns the env var name intentionally; the runtime resolves it downstream. Not a regression.

greptile-apps · 2026-02-02T07:38:46Z

docs/providers/azure-ai-foundry.md

+      "azure-ai-foundry-anthropic": {
+        baseUrl: "https://your-resource.openai.azure.com/anthropic",
+        apiKey: "${AZURE_AI_FOUNDRY_API_KEY}",
+        api: "anthropic-messages",


The example uses an apiKey value that looks like an env-var interpolation. The runtime normalizer strips that wrapper and stores just the env var name, which is fine if users go through the normal models.json generation path; but if someone copies this into a context expecting a literal key value, it may be confusing. Consider adding a short note that this apiKey form is treated as an env-var reference and normalized accordingly.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

Prompt To Fix With AI

This is a comment left during a code review. Path: docs/providers/azure-ai-foundry.md Line: 98:101 Comment: The example uses an `apiKey` value that looks like an env-var interpolation. The runtime normalizer strips that wrapper and stores just the env var name, which is fine if users go through the normal `models.json` generation path; but if someone copies this into a context expecting a literal key value, it may be confusing. Consider adding a short note that this `apiKey` form is treated as an env-var reference and normalized accordingly. <sub>Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!</sub> How can I resolve this? If you propose a fix, please make it concise.

Same pattern as the Venice provider docs (line 243). This is the standard convention for env-var references in explicit config examples throughout the project.

Takhoffman · 2026-02-02T13:04:09Z

Hello, have you tested this with some of the other models that azure provides too such as deepseek?

timmoh · 2026-02-03T08:04:38Z

great work. @alexpwrd @Takhoffman is this coming to release soon?

shudonglin · 2026-02-04T08:25:45Z

when will this feature release? i am waiting for the Azure Foundry support as well.

davidhgd · 2026-02-05T11:17:39Z

Thank you @alexpwrd great work!

rafalzawadzki · 2026-02-05T11:29:46Z

really looking for this to get in
previous contributions to solve this were automatically closed despite the need

nightfullstar · 2026-02-09T19:03:22Z

Really looking for this to get in as well

sebslight · 2026-02-13T03:45:19Z

Closing as duplicate of #12059. If this is incorrect, comment and we can reopen.

schupat · 2026-02-13T07:31:21Z

@sebslight I don't get it, why close this one and not the #12059?
This one is open for almost two weeks the other one just a couple days.

Sockolet · 2026-02-17T08:05:50Z

Can this be reopened and merged, please?

openclaw-barnacle bot added docs Improvements or additions to documentation agents Agent runtime and tooling labels Feb 2, 2026

style: fix formatting in azure-ai-foundry docs

4e6d39d

greptile-apps bot reviewed Feb 2, 2026

View reviewed changes

Takhoffman self-assigned this Feb 2, 2026

surajssd mentioned this pull request Feb 10, 2026

feat(agents): Add Azure AI Foundry credential support #12059

Closed

Takhoffman removed their assignment Feb 12, 2026

sebslight closed this Feb 13, 2026

haxudev mentioned this pull request Mar 19, 2026

feat: Add Microsoft Foundry provider with Entra ID authentication #50299

Closed

20 tasks

Uh oh!

Conversation

alexpwrd commented Feb 2, 2026 • edited by greptile-apps bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Notes

Test plan

Greptile Overview

Greptile Summary

Confidence Score: 2/5

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 2, 2026

Choose a reason for hiding this comment

Uh oh!

alexpwrd Feb 2, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 2, 2026

Choose a reason for hiding this comment

Uh oh!

alexpwrd Feb 2, 2026

Choose a reason for hiding this comment

Uh oh!

Takhoffman commented Feb 2, 2026

Uh oh!

timmoh commented Feb 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

shudonglin commented Feb 4, 2026

Uh oh!

davidhgd commented Feb 5, 2026

Uh oh!

rafalzawadzki commented Feb 5, 2026

Uh oh!

nightfullstar commented Feb 9, 2026

Uh oh!

sebslight commented Feb 13, 2026

Uh oh!

schupat commented Feb 13, 2026

Uh oh!

Sockolet commented Feb 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

10 participants

alexpwrd commented Feb 2, 2026 •

edited by greptile-apps bot

Loading

timmoh commented Feb 3, 2026 •

edited

Loading