Azure AI Search

Managed cloud search and retrieval for agent memory. Manage indexes, ingest documents, and run keyword, vector, and hybrid search to ground RAG pipelines in your own content.

14 actions available

New or changed content needs to reach the search index

Agent reads the content and any vector embedding from upstream

Agent writes with Index Documents using a mergeOrUpload action

Agent confirms the write with Count Documents and Get Index Statistics

Agent runs Search Documents to confirm the content is retrievable

Agent reports the index size change to the owning team

Any Delete Index call waits for an owner to approve

What This Integration Enables

The Azure AI Search connector covers the full index lifecycle. Agents create or update an index with its field schema and optional vector and semantic configurations, ingest documents in batches with per-document upload, merge, or delete verbs, and query with the flagship Search Documents action. That action runs keyword, vector, or hybrid search with filtering, faceting, paging, and optional semantic ranking. Suggest and Autocomplete power search-as-you-type. For pull-model ingestion, agents list indexers, trigger a run, and monitor status. Least-privileged query keys cover the read operations; admin keys are required only for writes. Agents typically vectorize content with a provider such as [Azure OpenAI](/integrations/azure-openai) or [OpenAI](/integrations/openai-ai) before an Index Documents call, then rerank retrieved candidates with [Cohere](/integrations/cohere) before presenting the top results.

Without FlowRunner

Content not searchable Material lives in stores no agent can query for grounding

Manual reindexing Someone triggers ingestion by hand when content changes

Single-mode search Keyword-only search misses semantically related content

With FlowRunner

Grounded retrieval Keyword, vector, and hybrid search feed an LLM its context

Ingestion on schedule Index Documents and indexers keep the index current

Ranked by relevance Semantic ranking reorders results by meaning, not just terms

Use Case Scenarios

Hybrid RAG Over Your Content

A team wants an assistant grounded in its own documents. The agent embeds each document with an external model and writes it with Index Documents using mergeOrUpload, so re-running never creates duplicates. At query time it embeds the user question, passes it to Search Documents as a hybrid query, and adds semantic ranking. The retrieved documents become grounding context for an LLM. Keyword recall and vector recall combine in one call.

Search-As-You-Type

A product surface needs type-ahead. With a suggester defined on the index, the agent calls Suggest to return matching documents as the user types and Autocomplete to complete query terms. The experience runs against the same index that powers full search, so there is one source of truth for what is searchable.

Scheduled Indexer Runs

Content lives in a data source an indexer can pull from. On a schedule the agent calls Run Indexer, then polls Get Indexer Status until the run finishes, and reports items processed and any failures to the owning team. Ingestion happens without anyone opening the Azure portal.

Human-in-Loop Highlight

Searching an index is read-only. Deleting one erases every document it holds. [Human-in-the-loop](/concepts/human-in-the-loop/) is an execution pattern where AI agents pause autonomously, assemble the relevant context and the decision choices available, route to a human via their preferred channel, and resume the moment the human responds. The Azure AI Search connector places that pause on Delete Index, the one operation that cannot be undone. When a flow reaches a Delete Index step, the agent first pulls the document count and storage size with Get Index Statistics, then asks the owner through their channel: "The product-docs index holds 51,000 documents and backs live search on the help site. Deleting it is permanent. Delete it?" The agent handles ingestion, querying, and indexer runs on its own. A person owns the decision to remove an index the business is serving from.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via preferred channel

Human decides

Agent resumes with decision

Agent Capabilities

14 actions

Indexes

Create Index Create or update a search index by name, defining the field schema plus optional vector search and semantic configurations. Idempotent PUT.
List Indexes List all search indexes defined on the service, with their fields and configuration.
Get Index Retrieve the full definition of a single index by name, including fields, analyzers, vector search, and semantic configuration.
Delete Index Permanently delete an index and all of its documents. This cannot be undone.
Get Index Statistics Return the document count and storage size for an index.

Documents

Search Documents The flagship query operation. Run keyword, vector, or hybrid search with filtering, faceting, paging, and optional semantic ranking. Use it to retrieve grounding context for an LLM.
Index Documents Batch upload, merge, or delete documents in an index. Each entry carries a @search.action verb plus the document fields.
Get Document Retrieve a single document from an index by its key value.
Count Documents Return the current number of documents in an index.
Suggest Return type-ahead document suggestions from a configured suggester for search-as-you-type experiences.
Autocomplete Return completed query terms for a partial query from a configured suggester.

Indexers

List Indexers List all indexers defined on the service. Indexers automate pull-model ingestion from a data source into an index.
Run Indexer Trigger an on-demand indexer run that reads from its data source and updates the target index.
Get Indexer Status Return the current status and execution history of an indexer, including the last run result and any errors.

Start building with Azure AI Search

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us