Context provider integrations

This component groups external integrations that supply, transform, retrieve, or persist context around an agent invocation. Most attach through the context-provider abstraction. A provider page can also include a provider-hosted tool when that is the external system's implementation of the same context pattern; those pages identify the mechanism explicitly.

Organizing by the framework integration surface keeps one provider page for its related storage, memory, RAG, pre-processing, or CodeAct patterns instead of duplicating the provider across feature categories. Use the Common patterns column below to browse by the outcome you need.

The documentation distinguishes several common patterns:

• Conversation storage reloads and persists the exact message transcript.
• Memory extracts and recalls selected durable knowledge from prior interactions.
• RAG retrieves relevant information from an external knowledge source.
• Pre-processing transforms incoming files or other content before model invocation.
• CodeAct contributes a code-execution tool and manages the execution environment.

These patterns describe common uses, not hard limits. A provider can combine multiple patterns or implement a different behavior entirely.

For the lifecycle, built-in abstractions, and custom-provider implementation guidance, see Context provider concepts.

Available integrations

Provider	Common patterns	C#	Python	Go
Azure AI Search	RAG	✅	✅	❌
Azure Content Understanding	Pre-processing	❌	✅	❌
Azure Cosmos DB	Conversation storage; memory	✅	✅	❌
Hyperlight	CodeAct	✅	✅	❌
Local (.NET)	CodeAct	✅	❌	❌
Mem0	Memory	❌	✅	❌
Microsoft Foundry	RAG; memory	✅	✅	❌
Monty	CodeAct	❌	✅	❌
Neo4j	RAG; memory	✅	✅	❌
Redis	RAG; conversation storage; memory	✅	✅	❌
Valkey	Conversation storage	✅	❌	❌

Next steps

Learn how context providers work