PipesHub, Self-Hosted Workplace AI Search Across Google, Slack, and Jira

Summary

PipesHub is an open-source, self-hostable workplace AI platform that unifies search across Google Workspace, Microsoft 365, Slack, Jira, Confluence and more using knowledge graphs and page ranking for explainable results with citations. It supports BYO models, real-time or scheduled indexing, source-level access control, and a no-code interface for building custom AI agents.

Key Insight

• Unlike most enterprise search tools that are SaaS-only, PipesHub can be fully self-hosted via Docker Compose, making it viable for privacy-conscious setups or on-premise deployments
• Uses a knowledge graph backbone (Neo4j or ArangoDB) combined with vector search (Qdrant) for hybrid retrieval - this is a more robust architecture than pure vector search alone, providing explainability through proper citations
• Supports a wide range of document types including scanned PDFs (via OCRmyPDF), audio, video, images, and all common office formats
• Built on a solid stack: FastAPI + LangChain/LangGraph for LLM orchestration, Kafka for event streaming, Celery for task queues - indicating production-grade architecture rather than a weekend project
• MCP server compatibility means it can plug into Claude Desktop or other MCP clients as a tool, enabling AI assistants to search internal company knowledge
• The “bring your own model” approach means no vendor lock-in for the LLM layer