Skip to content
User guide

MCP servers, skills and context

Register or deploy MCP servers for your agents, keep reusable skills, and give AI your documents through knowledge collections.

MCP servers

AI > MCP servers is the catalogue of MCP servers your organisation can use: REM CloudHub's own, and the ones you add.

Register a server that already runs somewhere: give it a Name, the Organisation and Team, its Address (https://mcp.example.com/mcp), its Transport and Sign-in, the Credential as a secret name in the organisation's Vault (callers never see it), Who may use it and a Reason. Click Register server.

Deploy one on REM CloudHub: start from a Template, an Image (ghcr.io/acme/mcp-server:1.0) or a Repository and Branch; say How it talks (a Command such as node dist/index.js for stdio, or a Port and Path such as 8000 and /mcp), an optional Install step, Environment (NAME=value per line, not for secrets) and Secrets to ask for. Review, then Deploy. It runs as a project in your team and is also reachable at /mcp/g/<organisation>/<name>.

Each server's page lets you switch single tools off, attach the server to an agent ("It shows on the agent's page"), and see its health and usage.

Your own MCP servers (Settings > My MCP servers): add a server only you use, with no sign-in, a bearer token, a header or OAuth sign-in. Its token is kept in your Vault and renewed; sign out revokes it.

Skills

AI > Skills holds versioned instructions agents can load: write one in the editor or upload it, run its publish checks, and publish a version. Levels and approvals decide who may use a skill. Agents list the skills they read on their Knowledge and permissions card.

Context: knowledge collections

AI > Context gives AI your documents.

  • Collections: sets of documents the AI can read: uploaded files, text, and what sources bring in. Open a collection to Upload a file, Add text, see Versions, and set Collection access (who may read or change it).

  • Ask and search: ask a question and get an answer with its sources, find passages, or see the block of context an agent would be given. Rate an answer Good answer or Poor answer.

  • Sources: places the engine reads from on a schedule (a drive, a site, a database) into a collection; Connectors (AI > Connectors) are the plugins that bring that data in.

  • Memories: short facts the AI keeps about you and your work and reads with every question; Forget all clears them.

  • Conversations: your chats with your collections, kept with their sources.

  • API keys: keys starting cck_ that let a program search and add to collections as you.

In Genie (the assistant in the console) you can attach a collection, a database or a file area to a chat in one click; answers cite it, and if you lose access to it, the chat loses it too.

Download: PDF · Word · Markdown

MCP servers, skills and context | REM CloudHub docs · REM CloudHub