pipeshub-ai
PipesHub is an open-source platform for connecting AI applications to the knowledge stored across a company's business systems. It connects Slack, Google Drive, GitHub, Microsoft 365, Notion and more than 50 other systems, lets people search across them and get answers with block-level citations, and filters everything by what each person is allowed to see. The same context can be given to your own agents and coding assistants.
It suits organisations where knowledge is scattered across many tools and access control cannot be an afterthought. Permissions are resolved when a query runs, against the source system's own rules, and an agent connects as a specific person, so it retrieves exactly what that person may see. Beyond search it offers knowledge-graph retrieval, multimodal understanding of images, diagrams and scanned files (using Docling and pdfplumber, or a multimodal model for scanned PDFs), deep research, no-code agents and a code-execution sandbox for reports and charts. An MCP server and SDKs let tools such as Claude Code, Cursor or Codex query company knowledge, and you can bring any LLM provider and deploy it inside your own network.
Getting started: Docker with Compose v2 is required. Run `curl -fsSL https://get.pipeshub.com/install | bash` (the README also shows how to download and read the script first); the installer checks RAM and disk, lets you choose a slim or full deployment, generates random secrets and starts the stack, after which you open http://localhost:3000. Use an HTTPS endpoint if you deploy on a cloud server.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-11-07
- Commits, last 90 days
- 100+
- Releases, last 90 days
- 7
- Latest release
- v1.0.0 (2026-10-27)
- Open issues
- 222
- Recent issues with at least one comment
- —
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #2946 Stock MCP_SCOPES omits semantic:read and conversation:read, so they cannot be minted23 comments · 2026-08-11
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Who it fits
- Knowledge is spread across Slack, Drive, GitHub, Microsoft 365, Notion and other tools, and you want one search with citations.
- Answers must respect each person's access: permissions are checked against the source system at query time.
- You want to self-host with your own LLM provider and give the same context to agents through MCP.
- You only need one app such as Jira or Slack: a single MCP server is far lighter.
- You cannot run several services (graph database, message broker, key-value store) on a properly sized server.
- You want a managed cloud version: the README lists it as coming soon.
How it differs
The Slack MCP server reads one workspace live through Slack's API. PipesHub indexes many systems ahead of time and adds permission-aware, cited answers.
mcp-atlassian acts on Jira and Confluence: search, create and update. PipesHub is for finding answers across systems, not for editing them.
Compatibility
Quick start
Use cases
Alternatives
Featured in these lists
Comments
No comments yet. Be the first!