Azure AI Services
Scope and Handoffs
| Request Scope | Route To |
|---|---|
| Foundry projects, agents (Prompt/Hosted/Voice), model deployments, fine-tuning, quota, project RBAC, Foundry VNet | microsoft-foundry |
| Foundry IQ knowledge bases and connectors | foundry-iq (separate foundry-iq-skills plugin; if absent, say so) |
Azure API Management AI Gateway policies (llm-content-safety, semantic cache, token limits) |
azure-aigateway |
| Classic AI Search, Speech SDK/REST/batch, Document Intelligence, Content Understanding, Translator, standalone Content Safety, OpenAI SDK | Stay in azure-ai |
| Roles, private endpoints, Studio storage or data handling for Document Intelligence or Content Understanding calls (including CU in the Foundry portal) | Stay in azure-ai |
Services and Diagnostic Workflows
Read the matching reference before diagnosing errors, validating local inputs, or collecting support evidence:
| Service | MCP / CLI | Preflight & Diagnostics |
|---|---|---|
| AI Search (classic indexes, vector/hybrid/semantic queries, indexers/skillsets) | azure__search (search_service_list, search_index_get, search_index_query), az search |
Index, query, and indexer diagnostics |
| Speech (STT, TTS, Fast/Batch Transcription, Batch Synthesis, SDK logs) | azure__speech (speech_stt_recognize, speech_tts_synthesize), az cognitiveservices |
Speech SDK client diagnostics | Audio & REST/batch preflight |
Document Intelligence (v4.0 Analyze, layout/OCR, custom extraction, classifiers, composed models) |
az cognitiveservices |
Input, v4 migration, and Analyze diagnostics | Custom models, labels, and confidence |
| Content Understanding (analyzers, CU Studio, model defaults, custom analyzers) | az cognitiveservices |
Setup, API versions, and defaults | Custom analyzers | Diagnostics |
| Access, network, and data for Document Intelligence and Content Understanding | azure__role (role_assignment_list), az role, az network |
Roles and auth | Studio storage | Private networking | Data handling | Evidence without content |
| Translator (Text & Document Translation, glossaries, Blob auth) | az cognitiveservices |
Document Translation & glossary preflight |
Content Safety (standalone text:analyze, image:analyze, text:shieldPrompt, blocklists) |
az cognitiveservices |
Standalone API preflight & diagnostics |
| OpenAI (direct SDK/API usage) | az cognitiveservices |
.NET SDK |
If Azure MCP is not enabled: In Copilot CLI or Claude Code, ask the user to run /mcp and enable Azure MCP; in other hosts, ask the user to enable Azure MCP through the host's MCP configuration. Offline local preflight scripts do not require Azure MCP or Azure sign-in.
For keyless authentication patterns across Azure AI SDKs, read Azure Authentication Best Practices.
SDK Quick References
- AI Search: Python | TypeScript | .NET
- OpenAI: .NET
- Vision: Python | Java
- Transcription: Python
- Translation: Python | TypeScript
- Document Intelligence: .NET | TypeScript
- Content Safety: Python | TypeScript | Java