Can AI agents automate Loom?

Loom is automatable today through Atlassian's official remote MCP server, for Loom workspaces that are linked to an Atlassian site. This record is Loom video messaging. It is not Jira, Confluence, Bitbucket, Compass, Trello, or Jira Service Management. The server is https://mcp.atlassian.com/v2/mcp. Loom permission groups are read_loom and write_loom, both available with OAuth 2.1 or an API token, under scopes read:loom:agent-interface and write:loom:agent-interface. Documented read tools include getLoomVideo (metadata, and optionally transcript, comments, AI briefs, meeting action items, or links), lists of owned and shared videos, comments, meeting action items, a time-limited MP4 download URL, and folders. Write tools cover the two-step upload and publish flow, title, visibility, sharing, comments, reactions, folders, and moving a video. Most Loom tools are reached through discovery rather than sent as primary tools. The public developer pages opened for this pass document the JavaScript record SDK and embed SDK (npm @loomhq/record-sdk and @loomhq/loom-embed), which embed recording and playback with a public app id and a domain allowlist. That SDK is not a workspace administration API, so the API path is unassessed. Starter includes a free workspace for up to 10 users, with 25 recordings and a 5-minute recording limit; Business and Enterprise are paid. https://www.loom.com/llms.txt tells agents to use the Atlassian remote MCP server, and the copy opened on 2026-10-02 still points at https://mcp.atlassian.com/v1/sse rather than the v2 URL. Computer-use viability is unassessed. MCP scores 7.

Best path today: mcp · Overall automatability: 7/10