PipesHub: open source context layer, fully self-hosted, bring your own model

We have been building PipesHub, an open source context layer you can run entirely on your own infrastructure. Apache 2.0, no cloud dependency. What we mean by a context layer: most company knowledge is spread across Drive, Slack, Notion, Jira, Confluence, SharePoint, Outlook, and a bunch of other tools. PipesHub brings that information together in one place while preserving permissions and keeping track of where everything came from. Once it’s indexed, search, chat, agents, MCP clients, or whatever else you’re building can use the same context instead of integrating with every source separately. You build the connection to your company’s knowledge once, rather than rebuilding the same data plumbing for every new AI tool. Why it's worth self-hosting rather than renting: • Data stays in your VPC or on-prem. Nothing leaves the network. • Bring your own model — OpenAI, Gemini, Claude, Ollama, or any OpenAI-compatible endpoint. Works quite well with even local models. • You control indexing, permissions and storage directly, including what gets indexed and when. What it does: • 40+ connectors - Slack, Google Drive, Notion, Confluence, Jira, Outlook, SharePoint, Dropbox and more • Hybrid search + knowledge graph retrieval • Agentic RAG for answers and workflows, with visual citations, reasoning and confidence scores • Tuned to say "not found" rather than guess, and the confidence score tells you when it's reaching • Deep research across mail, meetings and transcripts • Actions, not just answers — draft and send mail, schedule meetings, run workflows • MCP server, so you can point Claude Desktop, Cursor or VS Code at your own index • No-code agent builder, plus REST APIs and Python/TS/Go SDKs if you'd rather build directly • SSO (Google, Microsoft, OAuth) and Slack bot integration We built it to be something you can actually run inside your infra and extend as needed. Would really appreciate feedback from this community, especially around deployment, integrations, and real world use cases. GitHub: https://github.com/pipeshub-ai/pipeshub-ai