PipesHub：连接企业知识与AI智能体

The open-source context layer for AI agents. PipesHub turns your company's knowledge (Slack, Drive, Jira, GitHub, Microsoft 365 and 40+ c…

The Open-Source Workplace AI Platform Website · Docs · Discord · Roadmap Translations: Français · Deutsch · 简体中文 · 日本語 · Русский · עברית · 한국어 · Español · Português · Türkçe · Tiếng Việt · Italiano · Čeština PipesHub - Explainable & Extensible PipesHub is an open-source platform for connecting AI applications to the knowledge stored across your company's business systems. Connect Slack, Google Drive, GitHub, Microsoft 365, Notion and 50+ other systems. Search across them and get answers with citations, filtered by what each person is allowed to see — or give that same context to your own agents, workflows and MCP clients. Tip Deploy with a single command: curl -fsSL https://get.pipeshub.com/install | bash ▶ Watch the 2½-minute demo: Claude Code answering from Jira, Slack, GitHub and Drive, with each person's permissions What can you build with PipesHub? Don't just use PipesHub — build on it. Give Claude Code, Cursor, or Codex secure access to your company's knowledge → About ten minutes once PipesHub is running with data indexed. Mint a Personal Access Token (no admin needed), connect your assistant (one command for Claude Code, one config file for Cursor or Codex), and ask "why was the retry logic in the billing worker changed?" — it answers from the incident postmortem, the pull request, the chat thread, and the design doc, each cited, and only if you're allowed to see them. Want the same retrieval inside your own code, or behind a search box for your team? The SDK starter and search example cover both. Built something? Show us. PipesHub in Action Connectors Citations All Records Knowledge Search Features Answers you can trust 📝 Explainable Answers: PipesHub delivers grounded answers with precise block citations to the original documents. 🔒 Permission-Aware Search: Enforces source-level access controls so users only see what they're authorized to. 🧠 Bring Your Own Model, Fully Self-Hostable: Any LLM provider, deployed in your VPC — data never leaves your infrastructure. Connected to your systems 🔌 Enterprise Connectors: 50+ connectors with real-time and scheduled indexing out of the box. 🕸️ Knowledge Graph Retrieval: Graph-backed retrieval that captures relationships across enterprise data. 🎙️ Multimodal Support: Image, diagram, and scanned-file understanding plus voice-based interaction. What you can build 🔍 Unified Search, Deep Research, and Agents: Search, Q&A, deep research, web search, and AI agents on one context layer. 🤖 No-Code Agents and Actions: Build agents visually and execute actions across enterprise tools. 📊 Artifacts and Code Execution: Generate reports, charts, and dashboards in a safe execution sandbox. 🛠️ Developer-First and Extensible: APIs, SDKs, MCP tools, custom connectors, and independently scalable services. PipesHub Cloud Prefer a fully managed PipesHub without running your own infrastructure? Sign up for PipesHub Cloud: the same connectors, permission-aware search, and APIs, with nothing to install. Coding agents connect to https://app.pipeshub.com/mcp with a personal access token. See For coding agents. Connectors 🚀 Deployment Guide PipesHub can be run locally or deployed on any server using Docker Compose. The interactive installer handles all configuration — including secrets, graph DB, broker, and image tag selection — and generates a .env for you. HTTPS on cloud servers: If you deploy PipesHub on a cloud server, use an HTTPS endpoint. Browsers block certain requests over plain HTTP. Use Cloudflare, Nginx, or Traefik to terminate TLS. A white screen after HTTP-only deployment is typically caused by this restriction. ⚡ Quickstart (Recommended) Requires Docker with Compose v2. One command: curl -fsSL https://get.pipeshub.com/install | bash This downloads the deployment files for the latest release into ./pipeshub and launches the interactive installer. Open http://localhost:3000 once it finishes. Prefer to read before running? Download and inspect the script first: curl -fsSL https://get.pipeshub.com/install -o pipeshub-install.sh less pipeshub-install.sh # review it bash pipeshub-install.sh The installer will: Check Docker, RAM, and disk prerequisites Ask whether you want a slim or full deployment Let you optionally customise the graph DB, message broker, and KV store Generate randomised secrets and write a .env file Pull images and start the stack Wait for PipesHub to become healthy, verify it is reachable, and print the URL 🛠️ From a cloned repository (developers) To build from source, contribute, or pin the installer to your checkout: git clone https://github.com/pipeshub-ai/pipeshub-ai.git cd pipeshub-ai # Same installer, run from the repo root ./install.sh Building local images from source requires this cloned-repo path (./install.sh --build); the one-command installer above always uses prebuilt images. Installer options Flag Description -y / --yes Accept all defaults; skip interactive prompts (CI-friendly) --version TAG Pin a specific image tag, e.g. --version 0.7.0 --reconfigure Re-run the wizard and overwrite an existing .env --print-env-only Write .env and print the compose command without starting containers Advanced options: CI environment variables, slim vs. full deployment types, manual Compose profile usage, and local source builds are covered in Advanced Deployment Options. Build with PipesHub The built-in search experience is one way to use PipesHub. The same connected, permission-filtered context is available to your own agents and applications — over MCP for any compatible client, or through the SDKs when you are calling it from your own code. An agent connects as a specific person rather than as the application, so it retrieves exactly what that person is allowed to see. Access is resolved when the query runs, against the source system's own permissions, instead of being approximated at build time. Step-by-step tutorials for the most common builds — an MCP for your coding assistant, private enterprise search, and SDK starters — live in p