# Zoom MCP Connector for Claude

**URL:** <https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922>\
**Category:** API and Webhooks\
**Created:** [April 10, 2026, 10:00am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922 "2026-04-10T10:00:15Z")\
**Posts on this page:** 12\
**Page:** 1

<div class="post-metadata">

**Author:** ![Luke7](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/luke7/32/70316_2.png) [@Luke7](https://devforum.zoom.us/u/Luke7)\
**Post date:** [April 10, 2026, 10:00am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/1 "2026-04-10T10:00:15Z")

</div>

This is more related to the new MCP connector that has been released but I’m not sure where this ticket should be logged.

Basically, the MCP connector is unable to download vtt/txt files for meetings that DO have VTTs in the zoom portal. This is trivially easy via the api but is not possible via MCP despite the MCP being promoted as a tool that can download meeting resources etc.  
```  
GET [https://api.zoom.us/v2/meetings/{meetingUuid}/transcript](https://api.zoom.us/v2/meetings/%7BmeetingUuid%7D/transcript) # returns metadata including the download\_url  
GET {download\_url} — # the URL from above, fetches the actual VTT file content  
```

Is there any way this can be added? For us this is a major missing feature and, after having waited a long time for Zoom to implement an MCP connector, we’ve had to build our own. But it would be nice if the official one did this so that we don’t have to implement auth and MCP infra ourselves.

Thanks.

---

<div class="post-metadata">

**Author:** ![chunsiong.zoom](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/chunsiong.zoom/32/35488_2.png) [@chunsiong.zoom](https://devforum.zoom.us/u/chunsiong.zoom)\
**Post date:** [April 10, 2026, 11:58am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/2 "2026-04-10T11:58:55Z")

</div>

@Luke7 this should be working. There might be additional settings within [zoom.us](http://zoom.us) which you need to turn on. needs to allow users/host to retain the transcript. I also enabled the setting for “Allow users to retain transcripts generated by meeting summary for use by other AI Companion services”

 ![image](https://us1.discourse-cdn.com/flex016/uploads/zoomdeveloper/original/3X/0/d/0df925f4f93b1de35af658f10bcfe8dfe88e2ac6.png)

---

<div class="post-metadata">

**Author:** ![chunsiong.zoom](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/chunsiong.zoom/32/35488_2.png) [@chunsiong.zoom](https://devforum.zoom.us/u/chunsiong.zoom)\
**Post date:** [April 10, 2026, 11:59am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/3 "2026-04-10T11:59:30Z")

</div>

Here are more details on the app

> **[Using the Zoom for Claude app](https://support.zoom.com/hc/en/article?id=zm_kb&sysparm_article=KB0085220)**
>
> The Zoom for Claude app allows users to search their Zoom meetings using natural language, retrieve meeting

---

<div class="post-metadata">

**Author:** ![Luke7](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/luke7/32/70316_2.png) [@Luke7](https://devforum.zoom.us/u/Luke7)\
**Post date:** [April 10, 2026, 12:55pm UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/4 "2026-04-10T12:55:28Z")

</div>

> [@chunsiong.zoom](#):
>
> Allow users to retain transcripts generated by meeting summary for use by other AI Companion services

Hi Chun, thanks for looking into this. I think the issue may well be a permissions / config issue but I’m now wondering if this is actually a limitation of the MCP server.

Apologies, I don’t find the documentation very clear - there are lots of settings and some of the settings you listed above are not visible to me. Not sure if that is user config or jurisdictional thing?

Anyway, an example from today…

**Background**

I have a meeting which was **not** recorded - this is standard practice for us - but does have a VTT (.txt file) available in the [zoom.us](http://zoom.us) portal. I can download this vtt via the zoom website.

 ![image](https://us1.discourse-cdn.com/flex016/uploads/zoomdeveloper/original/3X/2/e/2ec48040d3ab02abd7bfdbb64be7bfb16851b320.jpeg)

However, when I try to use the MCP connector to get hold of this VTT, I see the following response..

> **Please find this zoom meeting from yesterday** …
> 
> _'s Zoom Meeting_
> 
> _Time: 9:25 AM – 9:30 AM UTC_
> 
> _Host:_
> 
> _Duration: ~5 minutes_
> 
> _Has AI summary: Yes_
> 
> **Perfect, please get the vtt transcript of that meeting.**
> 
> _The search results showed `has_recording: false` for that meeting, which means there’s no cloud recording — and without a recording, there won’t be a VTT transcript file either._
> 
> _The meeting was only about 5 minutes long (9:25–9:30 AM), which may be why it wasn’t recorded. However, it did show `has_summary: true`, so there is an AI-generated summary available._

For the record, I can **reliably** download the VTT via the Zoom API so the API does support this. I’m just not sure why the MCP is gated by the `has_recording` flag.

Thanks.

---

<div class="post-metadata">

**Author:** ![chunsiong.zoom](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/chunsiong.zoom/32/35488_2.png) [@chunsiong.zoom](https://devforum.zoom.us/u/chunsiong.zoom)\
**Post date:** [April 14, 2026, 3:09am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/5 "2026-04-14T03:09:39Z")

</div>

@Luke7 let me feedback to the team on this, thanks for flagging this up

---

<div class="post-metadata">

**Author:** ![EricDC](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/ericdc/32/74616_2.png) [@EricDC](https://devforum.zoom.us/u/EricDC)\
**Post date:** [May 8, 2026, 8:45am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/7 "2026-05-08T08:45:16Z")

</div>

Hey @chunsiong — Following up on Luke7’s report from Apr 10/13. We’re seeing the same issue two+ weeks later: meetings with AI-generated summaries and available transcripts are blocked by the `has_recording: false` gate in the MCP connector.

Like Luke, we can access these transcripts via the Zoom web portal and the Zoom API directly. The constraint appears to be MCP-specific, not a platform limitation.

Can you provide an update on the team’s feedback? Is this a known limitation, or is there a timeline for a fix?

Context: We’re piloting the MCP for Claude integration across our organization and this is blocking wider adoption — we have 50+ meetings/month and can’t reasonably cloud-record all of them just to enable Claude access.

Thanks.

---

<div class="post-metadata">

**Author:** ![Rob6](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/rob6/32/74477_2.png) [@Rob6](https://devforum.zoom.us/u/Rob6)\
**Post date:** [May 8, 2026, 8:47am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/8 "2026-05-08T08:47:24Z")

</div>

+1 — hitting this and a couple of related issues with the Zoom for Claude connector. Adding repro details in case it helps the team scope the fix.

**Environment**

- Zoom for Claude connector, accessed via Claude Code MCP namespace `mcp __claude_ai_Zoom_for_Claude__ *`
- Commercial Zoom (not Gov), AI Companion enabled, no cloud recording
- Meetings are AI-Companion-transcribed (transcripts visible in Zoom Workplace Hub → My notes → Transcript widget)

**Symptom 1 — `get_file_content` returns empty for every My Notes doc**

`search_zoom` with `entity_type: "zoom_doc"` and `doc_view: "notes"` returns a populated list of My Notes docs with valid `file_id` values, plausible `create_time` / `modify_time` deltas (~20–45 min between create and last modify, consistent with meeting duration), and working `docs.zoom.us/doc/<file_id>` links. The Zoom Workplace UI shows full content (note body and Transcript widget) for these docs.

But `get_file_content` returns `{"file_content": "\n", "file_name": "..."}` for every one of them. Tested across multiple notes from today and from earlier in the week — same result.

The tool description states “This tool works for both Docs and My Notes,” so this looks like a real bug, not an unsupported capability. Likely cause: the connector is hitting a synchronous content endpoint without going through the export job flow (`POST /docs/files/{fileId}/export` → poll `/docs/exports/{exportId}/status` → `GET /docs/files/{fileId}/content`), which appears to be required for note-shaped Workplace content.

**Symptom 2 — `search_meetings` returns empty even when meetings exist**

```auto
search_meetings(from=<this week start>, to=<today>) → {"meetings": [], "page_size": 30, "has_more": false}
search_meetings(no filters, page_size=30) → {"meetings": [], "has_more": false}
search_meetings(from=today, to=today+1, include_zoom_my_notes=true) → {"meetings": []}

```

Same account, same time window, `search_zoom(zoom_doc, doc_view: notes)` returns 10+ meeting note docs from today alone. So the meetings clearly exist; `search_meetings` just isn’t seeing them. This blocks `get_meeting_assets` and `get_recording_resource` because both need a meeting UUID and there’s no other way to enumerate them through the connector.

**Symptom 3 — chat search empty**

```auto
search_zoom(entity_type: chat, page_size: 5) → {"data": []}

```

Probably out of scope for this thread, but listing for completeness.

**What does work**

- `search_zoom` with `entity_type: "zoom_doc"` (any `doc_view`)
- `recordings_list` (returns empty, but plausibly genuine — no cloud recording on this account)

**Net effect**

The “hand the assistant my Zoom transcripts and let it summarize / answer questions” flow that’s the obvious value prop of this connector is currently impossible end-to-end on an AI-Companion-only account: there’s no path from “what meetings did I have” → “give me the transcript text.” Each broken link has a separate symptom but they compound.

Happy to share specific tool-call payloads privately if useful.

---

<div class="post-metadata">

**Author:** ![Luke7](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/luke7/32/70316_2.png) [@Luke7](https://devforum.zoom.us/u/Luke7)\
**Post date:** [May 19, 2026, 12:25pm UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/9 "2026-05-19T12:25:08Z")

</div>

Hi @chunsiong.zoom - any update on this one? As you can see from the other comments, it does represent a fairly big blocker to corporate roll-outs as this is probably one of the “big” features that users expect from the MCP integration.

---

<div class="post-metadata">

**Author:** ![Jase](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/jase/32/74882_2.png) [@Jase](https://devforum.zoom.us/u/Jase)\
**Post date:** [June 2, 2026, 3:10am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/10 "2026-06-02T03:10:29Z")

</div>

`get_file_content` returns “path param is empty: fileId” despite valid file IDs coming back from `search_zoom`.

---

<div class="post-metadata">

**Author:** ![AndrewThal](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/andrewthal/32/76171_2.png) [@AndrewThal](https://devforum.zoom.us/u/AndrewThal)\
**Post date:** [August 24, 2026, 4:36am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/11 "2026-08-24T04:36:25Z")

</div>

Following up: Zoom MCP tool calls are failing before returning transcript content with:

```text  
MCP error -32600: Tool has an output schema but did not return structured content  
```

This prevents transcript retrieval. Can you confirm whether this is a known issue and whether a workaround or fix is available?

\*Co-authored with openai-codex/gpt-5.6-terra\*

---

<div class="post-metadata">

**Author:** ![chunsiong.zoom](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/chunsiong.zoom/32/35488_2.png) [@chunsiong.zoom](https://devforum.zoom.us/u/chunsiong.zoom)\
**Post date:** [August 24, 2026, 7:18am UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/12 "2026-08-24T07:18:20Z")

</div>

@AndrewThal can you share how you can reproduce that error message?

---

<div class="post-metadata">

**Author:** ![AndrewThal](https://sea2.discourse-cdn.com/flex016/user_avatar/devforum.zoom.us/andrewthal/32/76171_2.png) [@AndrewThal](https://devforum.zoom.us/u/AndrewThal)\
**Post date:** [August 24, 2026, 2:45pm UTC](https://devforum.zoom.us/t/zoom-mcp-connector-for-claude/142922/13 "2026-08-24T14:45:33Z")

</div>

Repro context: this hits any MCP client on mcp Python SDK 1.12.4+ that enforces schema/structuredContent parity. The Zoom tools involved (search, get\_recording\_resource) declare an outputSchema. When the server response doesn’t populate structuredContent matching that schema — even though the underlying data exists and is fetchable via the Zoom web UI/REST API, per the has\_recording/empty-get\_file\_content reports upthread — a strict client raises -32600: Tool has a output schema but did not return structured content instead of surfacing the content.

Concrete steps: call search for a completed meeting with an available recording/transcript, then call get\_recording\_resource on the matched meeting UUID. The call errors before returning transcript text, rather than returning empty or gated content like the earlier reports in this thread.
