PipesHub – Open‑Source Workplace AI Platform

What it is – PipesHub is a self‑hosted platform that lets you connect the many data sources inside a company (Slack, Google Drive, GitHub, Microsoft 365, Notion, …) and expose that knowledge to AI‑powered applications. It builds a searchable knowledge graph, stores vector embeddings, and then lets an LLM generate answers that are grounded in the original documents with precise citations.

---

Core capabilities

Capability	What you get
Explainable answers	LLM‑generated replies include block‑level citations to the source file, so users can verify the information.
Permission‑aware search	Queries are filtered by the same access‑control rules each employee has in the source system – no data is shown to users who can’t see it.
Bring‑Your‑Own‑Model	Any LLM or embedding model can be used (OpenAI, Anthropic, Ollama, etc.). The platform is fully self‑hostable; data never leaves your VPC.
50+ enterprise connectors	Ready‑made sync for Slack, Google Drive, GitHub, GitLab, Microsoft 365, Notion and many others, with real‑time or scheduled indexing.
Graph‑backed retrieval	Relationships between entities are stored in Neo4j (or ArangoDB) and combined with vector similarity (Qdrant) for richer search.
Multimodal support	PDFs (including scanned), Office docs, images, and voice can be parsed; OCR is done via a multimodal LLM or Docling.
No‑code agent builder	Visual UI to create agents that can call APIs, run code, or perform actions across the connected tools without writing code.
SDKs & MCP	Python, TypeScript, and Go SDKs plus a Model‑Context‑Protocol (MCP) server let developers embed PipesHub in their own apps or AI assistants.

---

Typical use‑cases

1. Enterprise Q&A – Ask “Why was the retry logic in the billing worker changed?” and get an answer that pulls from incident reports, PRs, design docs, and chat threads, each with a citation.
2. Context for AI agents – Connect Claude Code, Cursor, or any MCP‑compatible assistant so it can act with up‑to‑date, permission‑filtered company knowledge.
3. Custom search portals – Build a web or internal tool that queries the same knowledge graph via the SDK or MCP.
4. Report generation – Use the sandboxed execution environment to create charts or dashboards from indexed data.

---

Getting started (quick‑install)

curl -fsSL https://get.pipeshub.com/install | bash

The script downloads Docker‑Compose files, runs an interactive wizard (choose slim or full deployment, graph DB, broker, etc.), creates a .env with random secrets, pulls the container images and starts the stack. After it finishes, open http://localhost:3000.

For developers who want to build from source:

git clone https://github.com/pipeshub-ai/pipeshub-ai.git
cd pipeshub-ai
./install.sh            # or ./install.sh --build to compile images locally

The installer also supports non‑interactive mode (-y) for CI pipelines.

---

Architecture at a glance

• Web UI – Next.js front‑end (search, chat, admin console).
• API layer – Node.js service handling accounts, permissions, and the MCP endpoint.
• Connectors & Indexers – Python workers that pull data from external systems, parse files (pdfplumber, Docling, etc.), create embeddings, and store them.
• Storage –
  • Graph DB: Neo4j (default) or ArangoDB for relationship queries.
  • Vector store: Qdrant for similarity search.
  • Document store: MongoDB + optional object storage.
  • Cache & messaging: Redis (or Kafka for larger deployments).
• LLM integration – Any hosted or local model supplies embeddings and generates the final answer.

---

Extensibility

• Custom connectors – Add a new source by implementing the Python connector interface.
• SDKs – Python (pipeshub-sdk), TypeScript (@pipeshub-ai/sdk), Go (pipeshub-sdk-go).
• MCP – Model‑Context‑Protocol server lets any MCP‑compatible client (e.g., Omnigent) retrieve context as a specific user.
• Deployments – Docker‑Compose for local/dev, Helm/Kubernetes manifests for production HA setups.

---

License & community

• Apache 2.0 – permissive open‑source license.
• Active Discord community, issue tracker, and contribution guide.
• Roadmap is public; recent milestones include no‑code agents, MCP support, and production‑grade Kubernetes deployment.

---

TL;DR

PipesHub is a fully open‑source, self‑hosted workplace AI platform that stitches together enterprise data sources, enforces per‑user permissions, and provides explainable, citation‑backed answers via any LLM you choose. It ships with Docker‑Compose deployment, a visual no‑code agent builder, and SDKs for Python, TypeScript, and Go, making it a solid foundation for building secure, enterprise‑grade AI assistants and search tools.