Azure AI skill: AI Search, Speech and Document Intelligence for coding agents

Source: microsoft/azure-skills · Maintainer: Microsoft · License: MIT · Checked: 26 October 2026

What is the Azure AI skill?

The Azure AI skill gives a coding agent a quick map of four Azure AI services, AI Search, Speech, Azure OpenAI and Document Intelligence, with the MCP tools and CLI commands to call and condensed SDK notes for writing code against them. It is for developers adding search, transcription, text-to-speech or document extraction to an app on Azure.

Its description reads: "Use for Azure AI: Search, Speech, OpenAI, Document Intelligence. Helps with search, vector/hybrid search, speech-to-text, text-to-speech, transcription, OCR." Its triggers include "AI Search," "vector search," "semantic search," "transcribe," "OCR" and "convert text to speech."

It ships in microsoft/azure-skills, the Azure Skills Plugin, which the README groups under "Work across data, AI, identity, and platform services." The README says the repo is synced automatically from microsoft/GitHub-Copilot-for-Azure. The SKILL.md frontmatter lists version 1.2.2 and author Microsoft.

As of 26 October 2026, its skills.sh page listed about 567,700 installs (snapshot, 26 October 2026).

What does it do?

The SKILL.md is short: a routing table, MCP commands, capability lists and links to reference files. It contains no scripts.

Services and how to reach them:

MCP commands it names:

• AI Search: search_index_list, search_index_get and search_query, to list indexes, read an index definition and query it.
• Speech: speech_transcribe and speech_synthesize.

These map to the Azure AI Search and Speech tools in the Azure MCP Server, though Microsoft Learn labels them differently: "Get search index" (which returns all indexes when no name is given), "Query search index," and Speech "Recognize" and "Synthesize." If Azure MCP is off, the skill tells the agent to ask you to run /mcp and enable it in Copilot CLI or Claude Code.

Capabilities it lists. For AI Search: full-text search with linguistic analysis, vector search, hybrid keyword plus vector search, and AI enrichment such as entity extraction and OCR. For Speech: real-time and batch transcription, neural voices with SSML, speaker diarization and custom models.

SDK quick references. Fourteen condensed files, about 30 lines each, with the install command, imports, non-obvious patterns and best practices:

The Python AI Search note, for example, recommends hybrid search, semantic ranking for natural language queries, batches of 100 to 1,000 documents, vector dimensions that match your embedding model, and creating suggesters when the index is created because they cannot be added later.

Auth guidance. A shared reference says to use managed identities and Azure RBAC in production and to reserve DefaultAzureCredential for local development, with per-language examples.

For deeper material the skill links out to Azure AI Search documentation and Azure AI Speech documentation.

How do you install the Azure AI skill?

The repo README documents these routes.

Claude Code: the Azure Skills Plugin. Run inside Claude Code:

/plugin install azure@claude-plugins-official

Or run /plugin and search for "azure". The plugin bundles all the Azure skills with the Azure MCP Server, which the README says gives agents "200+ structured tools across 40+ Azure services." Update with /plugin update azure@claude-plugins-official. See Claude Code plugins and Claude Code MCP.

GitHub Copilot CLI:

/plugin marketplace add microsoft/azure-skills
/plugin install azure@azure-skills

Codex and Cursor. The README says to run codex plugin marketplace add microsoft/azure-skills, then install azure from /plugins; for Cursor, use the Cursor Marketplace or Settings, then Plugins. See Codex skills and Cursor skills.

APM. The README also documents apm install microsoft/azure-skills for installing across several agents at once.

Skill only. The command on skills.sh:

npx skills add https://github.com/microsoft/azure-skills --skill azure-ai

This copies the skill files but not the MCP configuration. The SDK notes still work, but the search and speech tool calls need the Azure MCP Server set up separately.

Prerequisites, per the README: an Azure subscription, Node.js 18 or later for npx, and the Azure CLI signed in with az login. You also need the AI Search, Speech or other resources you want to work with.

How do you use it?

Name the service and, if you have them, the resource and index names. The skill is model-invoked, so the agent can load it when you mention AI Search, transcription, OCR and similar tasks.

Example prompts (suggestions):

• List the indexes on my search service srch-docs and show the fields in the products index.
• Query the products index for "waterproof hiking boots" and show the top five results.
• Add hybrid vector and keyword search to this Python API using azure-search-documents and managed identity.
• Transcribe meeting.wav to text with Azure Speech.
• Write a TypeScript function that extracts invoice totals with Document Intelligence.

For a retrieval-augmented app, see What is RAG for the overall pattern; this skill supplies the Azure AI Search side.

Tips and limits

From the repo and skills.sh, not from testing:

• It is a router more than a manual. The SKILL.md is brief and the SDK files are condensed. For index design, skillsets or agentic retrieval, the agent will still need Microsoft's docs.
• Coverage is uneven. Azure OpenAI and Document Intelligence have no MCP tools listed, and Azure OpenAI has a .NET reference only.
• Model and agent work lives elsewhere. Foundry projects, model deployments and hosted agents belong to the Microsoft Foundry skill.
• Queries use your credentials. MCP calls run as you, so the agent can read any index you can read.
• The plugin adds telemetry hooks. The Azure plugin's Claude Code hooks run a telemetry script on session start and after tool use. The README says to set AZURE_MCP_COLLECT_TELEMETRY=false to opt out.
• Security audits. On skills.sh, Gen Agent Trust Hub and Socket rated it Pass on 6 October 2026, and Snyk rated it Pass on 7 October 2026.

The MIT License lets you copy and change the skill as long as you keep Microsoft's copyright and license notices.

Save it in your Kipt library

If you want your own edited copy, for example with your search service, index names and field conventions written in, Kipt can hold it. Open Skills and paste the skill's GitHub folder link into the import box, or drop the folder or a .zip. A copy carries the skill files only, not the plugin's MCP configuration. Skills in your private GitHub repo install into Claude Code or Codex with one command:

npx skills add https://github.com/<you>/<repo>.git --all -g

Kipt is not affiliated with Microsoft.

Similar skills

• Microsoft Foundry skill: Foundry projects, model deployments and agents, from the same plugin.
• Azure prepare skill: plans and generates infrastructure when you are ready to host the app.
• Azure diagnostics skill: troubleshooting Azure apps in production.
• PDF skill: Anthropic's skill for reading and creating PDFs locally.
• MCP builder skill: building your own MCP server.