Pipeshub AI

What pipeshub-ai MCP server does

The pipeshub-ai MCP server connects MCP clients to an indexed knowledge layer for workplace systems. PipesHub lists Slack, Google Drive, GitHub, Microsoft 365, Notion, and more than 50 other connectors. It brings those sources into a common search experience instead of requiring an agent to query each service separately.

The server is intended for questions, retrieval, and agent workflows that need evidence from company data. Results can include citations pointing back to source documents, and access is filtered according to the permissions of the person making the request. This makes it suitable for internal research, support questions, codebase investigations, and other tasks where an answer should be traceable to source material.

PipesHub is open source and self-hostable. The README also describes SDKs, APIs, custom connectors, built-in agents, and independently scalable services, so the MCP interface is one way to use the broader platform.

How it works

PipesHub connects to business systems through enterprise connectors and indexes their content. The README describes both real-time and scheduled indexing, along with graph-backed retrieval that captures relationships across enterprise data. A client sends requests through MCP; PipesHub searches the connected knowledge and returns relevant context or answers with citations.

Permission enforcement happens at the source level. An agent does not receive records that the requesting user is not allowed to see. This is useful when a question spans several systems, such as a project discussion in Slack, an issue in Jira, a code change in GitHub, and related documents in Drive.

The platform can be deployed with an LLM provider of the operator's choice and can run inside the operator's infrastructure. The README says it supports multimodal content, including images, diagrams, scanned files, and voice interaction, although the specific behavior available through MCP is not detailed in the provided material.

Setup and configuration

The pipeshub-ai MCP server is used after PipesHub is running and data has been indexed. The recommended deployment method requires Docker with Compose v2. The provided installer can be fetched with:

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

The installer downloads deployment files into a pipeshub directory, launches an interactive setup process, and generates a .env file. It handles configuration such as secrets, the graph database, the broker, and image tag selection. The local interface is available at http://localhost:3000 after setup completes.

For a cloud-server deployment, the README recommends an HTTPS endpoint and notes that Cloudflare, Nginx, or Traefik can terminate TLS. Once the platform is running with indexed data, create a Personal Access Token and configure the MCP client. The repository specifically documents a one-command connection for Claude Code and configuration-file connections for Cursor and Codex.

Tools and capabilities

The pipeshub-ai MCP server supports these documented capabilities:

• Search across connected enterprise knowledge sources.
• Return answers grounded in indexed content with citations.
• Enforce source-level permissions during retrieval.
• Provide context to MCP clients, agents, and workflows.
• Use connectors for Slack, Google Drive, GitHub, Microsoft 365, Notion, and 50+ other systems.
• Support self-hosted deployment with a chosen LLM provider.

Limitations and notes

The provided material does not list individual MCP tool names, request schemas, transport details, or a complete client configuration example. Connector availability may also depend on the PipesHub deployment and its indexing configuration. PipesHub Cloud is described as coming soon, so the documented deployment path is self-hosted installation.

The README says HTTPS is important for cloud deployments because browsers can block some requests over plain HTTP. The platform also includes capabilities beyond MCP, such as no-code agents, actions, artifacts, code execution, and web search; their availability and configuration through this server are not specified here.