Integrate with Enterprise Systems
Connect Haystack Enterprise Platform to the systems you use every day. Combine pipeline components, agent tools, MCP, and data sync options to reach your systems. This guide explains how to configure common connections.
Connect to Microsoft SharePoint and Microsoft 365
For content in SharePoint sites or libraries, use the Microsoft SharePoint integration components:
- Indexing pipelines: MSSharePointFetcher downloads file content for embedding and storage. Pair it with MSSharePointRetriever when you search SharePoint first, then fetch the matched items.
- Query pipelines: MSSharePointRetriever searches SharePoint at query time. Connect MSSharePointFetcher downstream when you need full file content in the answer path.
Connect Microsoft credentials through workspace or organization secrets and integrations, following the same pattern as other hosted services in Secrets and Integrations.
Connect to Confluence, Jira, WSO2, and other SaaS
You can use one of the following approaches to connect to Confluence, Jira, WSO2, and other SaaS:
- Export or sync content into files or a database you control, then index with standard file converters, Snowflake, or an external database.
- Call vendor REST APIs from a custom component or from an Agent tool that wraps the API.
- Expose an MCP server that wraps the vendor API and attach it as an Agent MCP tool. For details on MCP tools, see Agent Tools.
Amazon Bedrock Knowledge Base retrievers can return documents whose metadata mentions Confluence or SharePoint as a source type when that knowledge base is configured in AWS. That path is AWS-side configuration, not a direct Confluence connector in Haystack Enterprise Platform.
Connect to HTTP and OpenAPI Services
Agents can call external HTTP APIs when you define tools that wrap those calls. For OpenAPI-described services, use the platform's OpenAPI tool patterns in Agent Tools and the OpenAPI integration components in the reference library.
For synchronous pipeline integration (non-agent), implement a custom component or use an existing HTTP-based component from the reference library.
Use Tools through the Model Context Protocol (MCP)
MCP lets agents use tools hosted on external MCP servers. You can also expose deployed pipelines as tools on the workspace MCP server for other clients. For details, see Model Context Protocol and Using deepset MCP Server.
Configure Bulk and Scheduled Data Movement
To keep your indexes fresh:
- Upload files through the UI or REST API. For details, see Upload Files.
- Run indexing pipelines on a schedule outside the platform (for example, CI or your orchestrator) using the API or SDK.
Choose an Integration Pattern
| Goal | Pattern |
|---|---|
| Search over SharePoint documents in an index | SharePoint retriever + fetcher, then document store in an indexing pipeline |
| Search SharePoint at query time | SharePoint retriever (and fetcher if needed) in a query pipeline |
| Agent looks up live tickets or records | Python tool or MCP server over the vendor REST API |
| Periodic warehouse sync | ETL into storage, then index |
| Call your internal microservice | OpenAPI or HTTP tool on an Agent |
Related Information
Was this page helpful?