# PipesHub

> Open-source workplace AI platform that connects enterprise knowledge from Slack, Google Drive, GitHub, Microsoft 365, and 50+ systems to AI agents with permission-aware search and verified citations.

PipesHub is an open-source (Apache 2.0) workplace AI platform that securely connects a company's knowledge across business systems to AI agents and search. It is self-hostable via Docker Compose with a single install command, and a managed cloud edition is listed as coming soon with a waitlist available.

## What It Is

PipesHub sits between a company's existing tools—Slack, Google Drive, GitHub, Microsoft 365, Notion, Jira, Salesforce, and 50+ others—and the AI agents or search interfaces that need to reason over that data. Rather than copying data into a vendor's cloud, PipesHub indexes it in your own infrastructure, enforces the source system's original access controls, and returns answers with precise block-level citations back to the original documents. The platform is built around three pillars the project describes as "Trust It, Own It, Build On It."

## How Retrieval Works

PipesHub uses a knowledge graph-first retrieval architecture rather than pure vector similarity search. The system combines a graph database (Neo4j by default, or ArangoDB) with a vector store (Qdrant) to capture relationships across enterprise data, not just keyword or embedding proximity. The project describes this as "Knowledge Graph + Page Ranking" to deliver structured reasoning across documents, entities, and metadata. Answers include a confidence level, citation count, and direct links to source paragraphs or page numbers.

## Permission-Aware by Design

A core design constraint is that access controls from the source system are inherited and enforced at query time. If a user cannot see a document in Google Drive, they cannot find it through PipesHub search either. Agents connect as a specific person rather than as an application, so they retrieve exactly what that person is authorized to see. The enterprise edition adds RBAC, SSO + SCIM, multi-tenancy, audit logs, and agent/connector governance on top of this baseline.

## Tech Stack and Deployment Model

The platform has three main layers: a Next.js web app for search, chat, and admin; a Node.js API for accounts, permissions, and knowledge bases; and Python services for connectors, indexing, and query answering. Data is stored in a knowledge graph, Qdrant for vectors, and MongoDB. Redis handles caching and local inter-service messaging; Kafka is used in larger deployments. The project is model-agnostic—users bring their own keys from OpenAI, Anthropic, Google, or Mistral, or run fully local with Ollama. SDKs are available for Python, TypeScript, and Go, and an MCP server is published separately for integration with MCP-compatible clients such as Claude and Cursor.

## What You Can Build On It

Beyond the built-in search and chat interface, PipesHub exposes the same permission-filtered context layer to custom agents and applications. The no-code agent builder lets teams create department-specific agents that can take actions across connected tools—creating Jira tickets from emails, syncing notes to Confluence, or triggering cross-system investigations. Artifacts and code execution support generating reports, charts, and dashboards in a sandboxed environment. The MCP server and REST APIs make the knowledge graph accessible to any compatible AI client.

## Update: v0.8.0

The latest release is v0.8.0, published in September 2026. The GitHub repository shows active development with commits every month. The roadmap (published on Notion) lists completed items including a no-code agent builder, MCP server and client support, developer SDKs, code search across GitHub and GitLab, and production Kubernetes deployment with HA defaults. Upcoming items include personalized search based on team and role, and PageRank-augmented relevance across the knowledge graph. A PipesHub Cloud managed offering is listed as coming soon with a waitlist.

## Features
- Permission-aware search with source-level access control enforcement
- Knowledge graph-first retrieval using Neo4j or ArangoDB + Qdrant
- Precise block-level citations with page numbers and paragraph references
- 50+ enterprise connectors with real-time and scheduled indexing
- No-code agent builder for department-specific AI agents
- Bring Your Own Model: OpenAI, Anthropic, Google, Mistral, or Ollama
- MCP server for integration with Claude, Cursor, VS Code, and other MCP clients
- Python, TypeScript, and Go SDKs
- Multimodal support: images, diagrams, scanned PDFs, and voice interaction
- Artifacts and code execution sandbox for reports, charts, and dashboards
- Single Docker Compose install command
- Production Kubernetes deployment with HA defaults
- SOC 2 Type I & II, ISO 27001, VAPT certified (enterprise)
- SSO + SCIM, RBAC, multi-tenancy, audit logs (enterprise)
- Self-hosted, VPC, on-premise, and air-gapped deployment options

## Integrations
Slack, Google Drive, GitHub, GitLab, Microsoft 365, Notion, Jira, Confluence, Salesforce, Gmail, Outlook, Google Calendar, Neo4j, ArangoDB, Qdrant, MongoDB, Redis, Kafka, Docker, Kubernetes, OpenAI, Anthropic Claude, Google Gemini, Mistral, Ollama, Claude (MCP), Cursor (MCP), VS Code (MCP)

## Platforms
LINUX, WEB, API, VSC_EXTENSION, DEVELOPER_SDK, CLI

## Pricing
Open Source, Free tier available

## Version
v0.8.0

## Links
- Website: https://pipeshub.com
- Documentation: https://docs.pipeshub.com
- Repository: https://github.com/pipeshub-ai/pipeshub-ai
- EveryDev.ai: https://www.everydev.ai/tools/pipeshub