Searching Loom with an AI agent
Teams use Loom for walkthroughs, standups, demos, and decisions. The Rubi Loom MCP lets Claude search those recordings and read transcripts, summaries, chapters, and AI-extracted action items.
The Model Context Protocol (MCP) server limits what the agent can access. The Rubi Odoo MCP docs explain the shared access pattern.
What the Loom MCP can do.
Ask what the team decided in a recording and get the answer, action items, and a link. Tools include search_videos, get_transcript, and get_summary. get_video_details returns metadata, transcript, chapters, summary, comments, and tasks in one call.
The connector exposes read tools today. Built-in write operations remain disabled pending security review. Any future write access would require your sign-off and be enabled tool by tool.
Two layers of security.
Microsoft Entra ID identifies the person using the connector. That person registers their Loom session through a separate browser-only step. Azure Key Vault stores the session credential, and the two identities refresh independently.
Hundreds of recordings become searchable and quotable in chat, without anyone handing over a Loom login.
Related MCP connectors
The Loom MCP shares its access model with the Rubi Odoo MCP and the Rubi 365 MCP. The same pattern applies to other systems with an API, and more connectors are in development.
Frequently asked questions.
The connector has no active tools for changing or deleting videos.
The AI receives neither the Loom session credential nor the user's login. Each person can access only the Loom data available to their own session.
Search across recordings by topic and use transcripts or summaries to recover decisions and action items.