No other tool connects these dots. Fathom and Otter just transcribe. HubSpot and Salesforce just log the call. Armbrain remembers who said what, matches it to client context, and surfaces it when it matters.
Quick Start
Connect Zoom from your Armbrain dashboard or from Claude:
From the dashboard (recommended):
- Go to hq.armbrain.io and open the Connectors tab
- Find Zoom in the connector list and click + Connect
- Approve the connection on the Zoom authorization screen
- Done - Armbrain starts discovering transcripts within 30 minutes
From Claude:
connectors action=connect source_type=zoomOpen the returned authorization URL in your browser, approve Armbrain's access, and return to Claude. Check status with connectors action=status.
Requirements
- Zoom cloud recording must be enabled. Zoom only generates transcripts for cloud recordings, not local recordings. Check Zoom Settings > Recording > Cloud recording.
- Audio transcript must be enabled. In the same settings area, ensure "Audio transcript" is toggled on under cloud recording options.
- A Zoom Pro, Business, or Enterprise plan. Cloud recording is not available on the free plan.
- The Zoom user approving OAuth must have cloud recording access on their account.
What Gets Ingested
- Cloud recording transcripts only. The connector downloads the VTT (WebVTT) transcript file from each meeting, strips timestamps and formatting, and ingests the plain text with the meeting topic as the title.
- Completed transcripts only. Meetings where Zoom is still processing the transcript are deferred and retried automatically on the next poll cycle.
- No audio or video files. Only the text transcript is ingested, not the MP4/M4A recording files. Your recordings stay in Zoom.
How It Works
The Zoom connector uses routing_mode: 'global' - a single Zoom connection covers all your client minds. You connect once; Armbrain figures out which client each meeting belongs to.
Discovery cycle:
- Default poll interval: 30 minutes
- On first connection, Armbrain looks back 7 days for existing transcripts
- After that, it checks for new recordings since the last poll (with a 30-minute safety overlap to catch stragglers)
- Pagination handles accounts with many recordings
Transcript processing:
- Discovers meetings with completed transcripts via the Zoom API
- Downloads the VTT transcript file
- Strips WEBVTT headers, cue numbers, timestamps, and NOTE blocks
- Formats clean text with the meeting topic as a heading
- Routes the content to the appropriate client mind based on context
- Retains the original VTT as an artifact for re-extraction
What Armbrain Can Do with Your Zoom Data
Armbrain requests a broad set of Zoom permissions because it uses your Zoom account as a relationship intelligence source, not just a transcript pipe. Here's what each capability enables:
Meeting Transcripts (live today)
Every cloud-recorded meeting with a completed transcript is automatically ingested into your Armbrain memory. Decisions, commitments, stakeholder preferences, and action items are extracted and stored in the right client mind. No manual entry, no copy-pasting from meeting notes.
What this means for you: Walk into any client meeting and ask "What did we decide about the rebrand timeline?" Armbrain finds it in last month's transcript, even if you never wrote it down.
AI-Powered Meeting Search (coming soon)
Search across your entire Zoom meeting history, Team Chat conversations, and Zoom Docs with natural language. Armbrain uses Zoom's AI Companion API to find relevant context you'd never think to look for.
What this means for you: Before a client meeting, Armbrain can search "What has the Acme team said about budget concerns?" across every conversation you've ever had with them - not just the ones you remembered to take notes on.
Calendar Intelligence (coming soon)
Armbrain reads your Zoom calendar to automatically match meetings to clients by attendee email and domain. No manual tagging required.
What this means for you:
- Morning briefings that say "You haven't met with Acme in 3 weeks - their Q4 planning starts next month"
- Meeting prep that pulls in the right client context without you switching minds first
- Engagement scoring that tracks how often you're meeting with each client
- Automatic detection of new stakeholders entering a client relationship
Meeting Assets (coming soon)
Capture shared content from meetings: whiteboards, shared screens, polls, and files. These get ingested alongside the transcript for richer context.
What this means for you: That whiteboard sketch of the campaign timeline? Armbrain captures it. The competitive analysis someone screen-shared? Stored with the meeting context.
Recording Analytics (coming soon)
Track whether clients watch shared recordings. View watch time and engagement metrics.
What this means for you: Send a strategy review recording to a client CEO. If they haven't watched it after 3 days, Armbrain surfaces that in your briefing so you can follow up.
Archive Access (coming soon)
Access recordings that have been moved to Zoom's archive storage. For long client engagements, this means Armbrain can reach back beyond the active recording window.
What this means for you: Six months into an engagement, a client asks "What did we agree on in the kickoff?" Armbrain finds it even if Zoom moved the recording to archive storage.
Privacy Controls (coming soon)
Delete transcripts from Zoom after ingestion for clients with data retention requirements. An "ingest then purge" flow for data-conscious organizations.
What this means for you: Enterprise clients with strict data policies can use Armbrain knowing that transcript data doesn't linger in multiple systems.
OAuth and Permissions
Armbrain requests read-only access to your Zoom account. It does not modify, delete, or create any Zoom data (the delete-transcript scope is reserved for the future opt-in privacy purge feature described above).
Scopes Requested
| Scope | Purpose |
|---|---|
cloud_recording:read:list_user_recordings | List your cloud recordings |
cloud_recording:read:list_recording_files | Get recording file details |
cloud_recording:read:meeting_transcript | Download transcript text |
cloud_recording:read:content | Access recording content and metadata |
cloud_recording:read:recording | View recording details |
cloud_recording:read:registrant | Track webinar registrations |
cloud_recording:read:registrant_status | Check registrant status |
cloud_recording:read:recording_analytics_details | Recording view analytics |
cloud_recording:update:recover_single_recording | Recover deleted recordings |
cloud_recording:update:recover_meeting_recordings | Recover deleted meetings |
cloud_recording:delete:meeting_transcript | Privacy purge (future, opt-in) |
archiving:read:list_archived_files | Access archived recordings |
ai_companion:read:ask | AI-powered meeting search |
ai_companion:read:search | Cross-channel content search |
calendar:read:list_calendar_lists | View calendar list |
calendar:read:event | View event details |
calendar:read:list_events | List calendar events |
meeting:read:search | Search meeting history |
meeting:read:assets | Access meeting shared content |
user:read | Basic user identity |
Data Handling
- Encrypted at rest. Zoom OAuth tokens are encrypted using AES-256-GCM before storage, with the encryption key held outside the database. Transcript content is stored in a database that is encrypted at rest (AES-256) by our infrastructure provider; the transcript text itself is not separately encrypted at the application layer. Per-customer isolation is enforced by application-layer query scoping (every query is filtered by your customer and mind IDs). PostgreSQL Row-Level Security policies are defined in the schema as defense-in-depth, with database-level enforcement being activated in stages.
- No audio or video. Armbrain never downloads, stores, or processes your actual recordings. Only transcript text and meeting metadata.
- No sharing across customers. Every query is scoped to your customer ID and mind ID, so one customer's Zoom data is never returned to another.
Troubleshooting
No recordings appearing:
Confirm cloud recording is enabled in Zoom settings and that meetings have actually completed (not still in progress). Zoom can take several minutes to generate transcripts after a meeting ends.
Transcripts not ingesting:
Zoom's transcript processing can lag behind the recording. The connector retries automatically. If a transcript is not ready, it retries on the next poll cycle (up to 12 times).
Auth errors after token expiry:
The connector automatically refreshes OAuth tokens before they expire. If the refresh token itself expires (rare - Zoom refresh tokens are long-lived), reconnect from the dashboard or with connectors action=connect source_type=zoom.
"Zoom OAuth is not configured" error:
This means the Armbrain server doesn't have Zoom credentials configured. Contact support to fix this server-side setting.
Disconnect
From the dashboard, click the disconnect button on the Zoom connector card. From Claude:
connectors action=disconnect connector_id=<id>This clears the stored Zoom credentials and stops future polling. Ingested memories remain preserved. Disconnecting does not delete past learnings.