Loom MCP Tools: What AI Agents Can Do With Your Recordings
Vorec Team · 2026-10-24 · About 9 min read
Loom's MCP tools let an AI agent connected to Atlassian MCP find Loom videos and meeting recordings, read their transcripts, action items and comments, and write back titles, visibility, sharing, timestamped comments and new uploads. They are Loom read and write tools inside Atlassian MCP (v2), not a separate server, and they only cover Loom workspaces linked to an Atlassian site. Atlassian announced them on 22 October 2026, and we checked its documentation on 24 October 2026.
A screen recording used to be the end of a task: you recorded it, someone watched it, and the video sat in a library. On 22 October 2026 Atlassian described a different job for it. In Showing what you mean in the age of AI, Loom's head of product announced that Loom now supports the Model Context Protocol, so that recordings remain a source of context for agents and tools after the person who recorded them has handed off the work.
In practice that means Loom tools inside Atlassian MCP: an AI agent connected to your Atlassian site can look up Loom videos and meeting recordings, pull their transcripts and action items, leave comments at a timestamp, and even upload and share a new video. This post sets out what Atlassian's own documentation lists, how access and credits work, and what we think it signals for anyone who records software for a living.
Checked on 24 October 2026 against three Atlassian pages: the announcement (22 October 2026), the Supported tools reference and Get started with the Atlassian MCP server. Tool names and limits can change; treat those pages as the source of truth.
Quick answers. What it is: Loom read and write tools inside Atlassian MCP (v2), not a separate server. What agents can read: metadata, transcripts, agent briefs, action items, comments, attached links and developer context logs for videos and meeting recordings, plus a temporary MP4 download link. What they can write: title and description, visibility, sharing, timestamped comments or reactions, folders, and new uploads. Requirement: the Loom workspace must be linked to an Atlassian site.
What Atlassian announced on 22 October
The announcement bundles four things under one idea, "show your intent once":
- Record for agent, a recording mode that turns a short narrated screen recording into a prompt for an agent. We covered it in detail in Loom video prompts.
- The new Loom for Mac, a rebuilt desktop app that Atlassian says records at 4K by default and adds a Selection tool to pin a UI element and a Magic pen for freehand annotation.
- Loom tools in Atlassian MCP, the subject of this post.
- Rovo in Loom, which Atlassian describes as rolling out now, and which it says can find a video or moment across a library, answer questions about what was said, draft the follow-up, or turn a walkthrough into written steps, with each answer tied to the recording and timestamp.
The announcement names the agents it expects to use the MCP tools: Claude, Cursor, VS Code, Codex and Rovo. The Get started page separately lists supported clients including OpenAI ChatGPT, Claude, Docker, GitHub Copilot CLI, Google Gemini and Amazon Quick Suite, and gives setup commands for Claude Code and Codex.
The Loom tools, as Atlassian documents them
Atlassian splits the Loom tools into two groups, each with its own OAuth scope. Both work with OAuth 2.1 or API token authentication. The Supported tools page adds one condition at the top: only Loom workspaces linked to an Atlassian site are available.
Read tools (`read_loom`, scope `read:loom:agent-interface`)
| Tool | What the docs say it does |
|---|---|
| getLoomObjectsContent (primary) | Gets metadata or content for one or more known Loom objects by ARI, share URL or embed URL. Can request transcripts, agent briefs, action items, comments, attached links or developer context logs |
| listLoomObjects | Browses Loom objects you own, that are shared with you, or that are visible in the workspace, with filters for scope, content type and creation date |
| getLoomVideoDownloadUrl | Returns a temporary signed URL to download the video as an MP4 |
| listLoomObjectFolders | Lists folders in the current workspace |
Loom objects, in Atlassian's wording, currently means async videos and meeting recordings.
Write tools (`write_loom`, scope `write:loom:agent-interface`)
| Tool | What the docs say it does |
|---|---|
| createLoomObjectUpload | Starts creating a Loom object from a local video file. Returns a temporary upload recipe; it doesn't transfer or publish anything itself |
| publishLoomObject | Completes the upload after the file has transferred, and returns the share URL |
| updateLoomObjectMetadata | Changes the title, the description, or both |
| updateLoomObjectVisibility | Sets visibility to OWNER, WORKSPACE or PUBLIC |
| shareLoomObject | Shares with specific Atlassian users, optionally with a message, without changing general visibility |
| addLoomInteraction | Adds a text comment or emoji reaction at a specific timestamp |
| createLoomObjectFolder | Creates a folder, optionally inside another folder or Loom space |
| moveLoomObjectToFolder | Moves a video or meeting recording into a folder |
| recoverLoomVideo | Re-processes a video that failed to process or is stuck |
Two details stand out. First, the read side goes well beyond the transcript. Our reading is that agent briefs and developer context logs relate to Record for agent, which the help article says can capture console and network logs in the Chrome extension; the tools page itself doesn't make that link. Second, the write side lets an agent publish a video it did not record in Loom, through the two-step upload, and change who can see it.
How to connect it
Atlassian says the Loom tools are built into Atlassian MCP, so there is no separate server to set up.
- Already on Atlassian MCP v2: the Loom tools appear automatically once your Loom workspace is linked to your Atlassian site.
- On v1: update or reinstall the Atlassian connector in your AI client. The upgrade guide says that on 1 March 2027 any existing v1 use will automatically start exposing v2 tools, and that incompatible clients will need to clear cached client IDs or credentials.
- Claude Code: the Get started page gives `claude mcp add --transport http atlassian https://mcp.atlassian.com/v2/mcp`, then you authenticate in a Claude Code session.
- Codex: `codex mcp add atlassian --url https://mcp.atlassian.com/v2/mcp`.
- MCP gateways: append `?tools=all` to the server URL to get a flat tool list instead of the discover-and-execute flow.
None of the three pages we checked says which Loom plans include the MCP tools. Record for agent itself is in open beta for Business + AI and Enterprise plans, according to the announcement. Check your plan before you build on it; our Loom pricing breakdown lists the tiers.
Permissions and credits
Permissions. Atlassian says actions run within the team's existing Loom permissions, and that Atlassian MCP uses OAuth 2.1 so that actions respect users' existing access controls. Its own advice on the Get started page is direct: "Use least privilege, review high-impact changes before confirming, and monitor audit logs for unusual activity."
That advice matters more here than for read-only tools, because `updateLoomObjectVisibility` can set a recording to PUBLIC. A recording of an internal tool can show customer names, settings or tokens on screen. If you grant `write:loom:agent-interface`, decide in advance which changes you will confirm by hand.
Credits. The Get started page says that when an AI client calls Atlassian MCP to retrieve data or generate insights, it consumes Rovo credits from the same shared pool your organisation uses for Rovo Chat, Studio, Agents and the Teamwork Graph, and that usage depends on how much data is pulled and how much reasoning is needed. The Supported tools page gives per-call notes for Teamwork Graph tools (most are 1-10 credits per call); it does not list a figure for the Loom tools, so we don't state one.
What it means for screen recordings
What follows is our reading, not Atlassian's.
A recording is becoming structured data, not just a video. The tools an agent reaches for first are the transcript, action items and agent brief, not the pixels. That rewards recordings that are easy to parse: a clear narration that says what is happening and why, one task per recording, and a title and description that say what the video covers. A rambling ten-minute take is as hard for an agent to use as it is for a person.
Agents can now be part of the publishing loop. With the upload, metadata, visibility and folder tools, an agent can do the filing work around a recording: name it, describe it, put it in the right folder and share it. Whether you let it do that unattended is a policy decision, and Atlassian's least-privilege advice is a good default.
Video and written steps are converging. Rovo in Loom turning a walkthrough into written steps is the same idea as turning a screen recording into a help article. The recording becomes the source, and the formats people actually read are derived from it.
If you are new to the protocol itself, our explainer on what the Model Context Protocol is covers the basics, and Playwright MCP shows the other direction: an agent operating a browser rather than reading recordings.
Where Vorec fits
Loom's tools let agents work with recordings after they exist. Vorec focuses on making the recording itself a finished tutorial. It has its own macOS recorder, and an AI agent can drive it for you through the Claude Code plugin, capturing locally for you to review before anything is uploaded. It then drafts narration matched to the workflow it captured and generates the voiceover, so nothing is spoken into a microphone. In the editor you can add smooth cursor motion, cursor-follow zoom and click-based auto-zoom, and freeze-sync can hold a frame to give an explanation time to finish. The same capture can also produce a written step-by-step guide with annotatable screenshots, and narration regenerates in supported languages on eligible plans. You can upload an existing recording instead if you already have one. The hosted Agent API is available on the Business plan; plans start at $9/month (Starter), and a free trial is available (current allowances on the pricing page).
Teams that need agents to search, comment on and file an existing Loom library should evaluate Loom's MCP tools directly.
FAQ
Does Loom have an MCP server?
Not a separate one. Loom tools are part of Atlassian MCP (v2), at `https://mcp.atlassian.com/v2/mcp`. They become available once your Loom workspace is linked to your Atlassian site.
Can an AI agent read a Loom transcript?
Yes, according to Atlassian's Supported tools page. The getLoomObjectsContent tool can return transcripts, along with agent briefs, action items, comments, attached links and developer context logs.
Can an agent upload a video to Loom?
The docs list a two-step upload: createLoomObjectUpload returns an upload recipe for a local video file, and publishLoomObject completes it after the file has transferred and returns the share URL.
Which agents work with Loom's MCP tools?
Atlassian's announcement names Claude, Cursor, VS Code, Codex and Rovo. The Atlassian MCP setup page says it supports any app with MCP support and gives setup steps for clients including Claude Code and Codex.
Does using Loom tools over MCP cost anything?
Atlassian says calls through Atlassian MCP consume Rovo credits from your organisation's shared pool, with usage based on data volume and reasoning depth. It doesn't publish a per-call figure for the Loom tools.
Is this the same as Loom video prompts?
No. Video prompts (Record for agent) turn a new recording into a prompt you hand to an agent. The MCP tools let an agent work with recordings that already exist. Atlassian announced them together on 22 October 2026.
Want the recording to narrate itself? Record with Vorec, or upload one you already have, and get a narrated tutorial plus a written guide. Start free. Trial includes up to 3 projects; exports carry a watermark.