5 MINUTE READ
Context, MCP & skills
Learn what Loom stores in Context Hub and what its native inventories do.
Manage workspace context
Context Hub is where you manage project-scoped knowledge, rules, skills, and approved memory. Use it for stable guidance that should be available across related work, and keep each entry focused and current. Review the context attached to a task before sending sensitive information to a model provider.
Loom also discovers native skills in supported Claude, Codex, and OpenCode skill locations. Discovery is a read-only inventory: it does not run skill scripts or plugin hooks. Results can be partial when a source reaches the scan limit, and names or metadata may need review before a skill can be used by Loom.
Understand the MCP inventory
The MCP settings inventory scans supported native configuration sources and lists server definitions, transport type, source, endpoint or executable, and safe header/environment variable names. Secret values are not shown.
A discovered definition does not mean the server is running, connected, authorized, or available to an agent. The inventory is read-only: Loom does not launch local MCP processes or contact remote servers from this page. It does not yet manage connections, show tool catalogs, or make these definitions available as Loom tools. Native configuration remains owned by its source runtime.
Review discovered skills
Skill discovery identifies supported native skill roots for Claude, Codex, and OpenCode. It reports the source and metadata so you can review what is present. A scan-limit notice means the list may be incomplete; an invalid-name notice means the skill does not meet Loom's naming rules.
Discovery is not execution or automatic installation. Review a skill's instructions and any scripts yourself before choosing to use it in an agent workflow.