Repository navigation
Description
Why
Betty cannot join Zoom as a web-client bot: Zoom shows "automated bots aren't allowed ... must use Zoom RTMS", and Zoom's developer docs now say the Meeting SDK "is reserved for human use cases and does not support bots or AI notetakers" and that apps joining meetings outside their account must be authorized from March 2, 2026 (OBF/ZAK tokens or RTMS). We accept that wall. This issue compares the ways Betty can still take part in a Zoom meeting without a Zoom app or RTMS and without evading Zoom's bot detection (no stealth browsers, no fingerprint spoofing, no anti-bot bypass), recommends one, and tracks its prototype.
Constraints from the meetings design that every option must keep: consent first (nothing from Betty before the consent notice, and the notice must be delivered, else leave), hive is the policy point (voice rules, rate limits, budget), and transcripts are erased upstream after hand-off.
Options
1. Phone dial-in (considered, not pursued)
Considered, not pursued (owner decision): Betty phoning in through a telephony provider needs a paid Zoom host (dial-in is not on the free plan), a telephony account (about $1.10 per meeting-hour on Twilio US), and could only speak its consent notice, never confirm it.
2. Desktop companion (recommended): capture the human's own Zoom client on their own machine
A small desktop app on the host's (or any invited human's) computer, the pattern Plaud Desktop ships: it notices a Zoom, Teams or Meet call starting, asks "Let Betty listen to this call?" with the consent reminder, captures the call app's audio (everyone else) and the microphone (the user), and streams both to the hive as a live meeting. Betty's existing pipeline (transcript, Spek notes, research, replies) consumes it; replies show in a small companion window and in the hive panel/Inbox. Spoken replies through a virtual mic are out of scope for v1.
- No virtual driver on macOS: ScreenCaptureKit captures one app's audio (
SCStreamConfiguration.capturesAudio,excludesCurrentProcessAudio; microphone viacaptureMicrophoneon macOS 15+), and Core Audio process taps (CATapDescription,AudioHardwareCreateProcessTap, macOS 14.2+,NSAudioCaptureUsageDescription) tap a single process. Windows has WASAPI per-process loopback; Linux has PipeWire per-node capture. BlackHole/VB-Cable are only needed for the later "Betty speaks through the host's mic" step. - Compliance: nothing joins the meeting and nothing touches Zoom's servers or bot detection; it is the human's own client audio on their own machine, like Plaud Desktop or Zoom's own local recording. The legal duty is the one Zoom's Terms (section 7) put on every user: comply with recording-consent laws. So the companion requires an explicit per-call "I have told the participants" attestation (versioned like the Inbox recorder's
consent_version), shows a persistent red "Betty is listening" indicator while capturing, offers the consent notice on the clipboard to paste into the meeting chat, and stops on one click. Participants are not notified by Zoom, so this attestation is the consent gate; the hive records who attested, when, and which notice version. - Privacy/security: audio leaves the machine only while a capture the user accepted is running, only to the user's own hive, over TLS, as raw 16 kHz mono PCM chunks. The companion authenticates with a per-user, revocable companion token minted from the dashboard (owner or read-write only; only its SHA-256 is stored, like the iPhone widget token) that is admitted on the companion routes only and acts as the minting user at their live role (a revoked or downgraded user stops working at once). Nothing is stored on the laptop beyond the token (in the OS keychain). Read-only users cannot stream.
- UX: works on the free Zoom account and on any platform the user's client can join (Zoom, Teams, Meet, Webex, Slack huddles); two channels give "you" vs "the call" attribution (the call side is one mixed stream, no per-person names). Echo: without headphones the mic picks up the speakers; v1 documents headphones and drops mic segments that duplicate a call segment.
- Cost: none (uses the existing transcription service). Effort: medium-high (desktop app, three capture modules over time).
- Distribution (owner decision): one Tauri codebase in a new repo
hivecommons/quorum-companion(Apache-2.0, DCO); macOS first, then Windows, then Linux. No Apple Developer ID: builds are ad-hoc signed (codesign -s -), shipped as a GitHub Release.dmg/.zipand a Homebrew cask inhivecommons/homebrew-hive. First launch needs System Settings, Privacy & Security, Open Anyway, then the Screen & System Audio Recording and Microphone permissions. Homebrew 5 removed--no-quarantine, so the cask docs show the Open Anyway step. Ad-hoc signatures have no stable designated requirement, so macOS may forget the audio-capture grant after each update; mitigation: keep the bundle id (dev.hivecommons.quorum-companion) and install path fixed, and sign releases with one self-signed code-signing certificate kept as a CI secret (a stable certificate leaf gives a stable designated requirement, so grants survive updates; Gatekeeper still asks for Open Anyway once). Until that secret exists, builds are plain ad-hoc and may re-ask after updates;tccutil reset ScreenCapture dev.hivecommons.quorum-companionclears a stale entry.
3. Zoom "Live on Custom Live Streaming Service" (RTMP)
The host clicks More, Live on Custom Live Streaming Service, and Zoom pushes the meeting to an RTMP URL; Betty would transcribe that stream.
- Account tier: custom live streaming needs a Pro, Business, Education or Enterprise host and an admin setting ("Allow livestreaming of meetings", "Custom Live Streaming Service"); not available on the free account.
- Compliance: fully sanctioned and visible (Zoom shows everyone the meeting is live). Listen-only: no chat, no voice, so Betty cannot post the consent notice or answer in the meeting.
- UX/effort: the host must start it every meeting; 10 to 30 s of stream latency; needs a public RTMP ingest endpoint for the hive (new exposed surface). Cost: Zoom Pro license plus ingest bandwidth.
Other documented paths
- Zoom RTMS / Marketplace app: Zoom's sanctioned route for AI notetakers; excluded by this issue's brief (needs a Zoom app).
- Zoom's own cloud recording/transcript then import: paid feature, after the meeting only; could feed the Inbox later.
Comparison
| Desktop companion | RTMP live stream | |
|---|---|---|
| Works with the owner's free Zoom account | Yes | No (Pro+) |
| Touches Zoom bot rules | No (user's own client) | No (Zoom feature) |
| Consent delivery | Per-call attestation + indicator + clipboard notice | None in-meeting |
| Live Betty Q&A | Yes, in companion window and panel (voice later) | Transcript only, 10-30 s lag |
| Speaker attribution | You vs the call | None |
| Cost | $0 | Zoom Pro + ingest |
| Needs an owner purchase | No | Yes (Zoom Pro) |
| Other platforms for free | Zoom, Teams, Meet, Webex, Slack | Zoom only |
Recommendation
Build option 2, the desktop companion, macOS first. It is the only option that works with the free account, needs no purchase, stays clear of Zoom's bot rules because nothing joins the meeting, and reuses Betty's live pipeline unchanged. Option 3 is not worth it: listen-only, paid, and a new public ingest.
How it fits what hive already has
The dashboard's ● Record ("Recording into the Inbox", docs/inbox-recording.md) already captures mic + a meeting tab's audio in Chromium via getDisplayMedia, behind a versioned consent notice, owner or read-write only. Its ingest (POST /api/inbox/recordings, ordered chunks?seq=N, finalize) is post-meeting: chunks must be WebM/MP4 media, are concatenated at finalize, and transcribed afterwards, so Betty cannot react live. The companion keeps Record's consent model and chunk protocol shape (ordered seq, 409 with expected_seq on a gap, idempotent resend) but streams to a live meeting route so Betty answers during the call; when the meeting ends it lands in the Inbox as a normal meeting item, so no separate Record upload is needed. Record stays the no-install option (Chromium tab audio only; it cannot hear the Zoom desktop client).
Streaming contract (companion to hive to Quorum)
Documented in quorum docs/contract.md (new companion platform) and hive docs/meetings.md:
POST /api/companion/meetings{title, app, consent_version, consent_attested: true}creates a live meeting on platformcompanion(hive calls Quorummeeting.join {platform:"companion"}), records who attested and when, returns{id, sample_rate: 16000, max_chunk_bytes}.POST /api/companion/meetings/{id}/audio?channel=call|mic&seq=Nbody = raw 16 kHz mono s16le PCM; per channel, in order; 409expected_seqon a gap; a resent stored chunk answers 200.GET /api/companion/meetings/{id}/replies?after=Nreturns Betty's lines for the window (the consent notice first).POST /api/companion/meetings/{id}/end; "betty leave" said in the call also ends it.- Auth:
X-Hive-Companion-Token(per user, revocable, companion routes only). On hosted hives the hub waves/api/companion/through to the spoke, which does the real check (same pattern as the widget path). - Quorum: the
companionadapter segments each channel with an energy VAD, transcribes with the shared service, and emits normalsegmentevents (speaker= the user formic, "Call" forcall);chat.postqueues the line for the companion window (confirmedonce the companion has fetched it);sayisplatform_unavailablein v1.
Prototype PR set
- quorum:
companionplatform (audio ingest, VAD, transcription, reply outbox) + contract + tests: ✨ companion platform: desktop companion audio ingest (0.3.0) quorum#32, released v0.3.0; hive pin 🌱 deps: quorum v0.3.0 (desktop companion platform) #18622 - hive:
companionmeeting platform, Quorum client calls,/api/companion/*routes, docs: ✨ meetings: desktop companion meetings (/api/companion/*) #18611 - hive: per-user companion token (
/companion/token), hub passthrough for/api/companion/: ✨ meetings: per-user desktop companion tokens and hub passthrough #18643 - new repo
hivecommons/quorum-companion: Tauri macOS v1 (call detection, consent prompt, ScreenCaptureKit capture of the call app + mic, streaming, replies window), ad-hoc signed CI builds: ✨ macOS v1: call detection, consent prompt, capture and streaming to the hive quorum-companion#2, released v0.1.0 - homebrew-hive:
brew install --cask hivecommons/hive/quorum-companion: ✨ cask: quorum-companion 0.1.0 homebrew-hive#336
Status (2026-11-17)
Decision: the desktop companion is the Zoom path. Phone dial-in and RTMP are closed options; nothing further is planned for them.
- The hive image carrying ✨ meetings: desktop companion meetings (/api/companion/*) #18611, 🌱 deps: quorum v0.3.0 (desktop companion platform) #18622 and ✨ meetings: per-user desktop companion tokens and hub passthrough #18643 is deployed on the project's own hive, and
quorum-companionv0.1.0 is released (Homebrew cask and GitHub Release). - The companion ingest path has been exercised end to end by an automated test client using synthetic audio (create meeting with consent attestation, stream 16 kHz PCM, read replies, end): meetings start, transcribe and end cleanly. That run measured Betty's answers, not the desktop app, so it does not replace the live test below.
Next
- Live test of the macOS app on a real Zoom call (needs a person with the app installed): [scanner] Desktop companion: live end-to-end test on macOS with a real Zoom call #18659
- Windows capture module (WASAPI process loopback): Verify Windows capture on a real call and publish Windows release assets (code merged in quorum-companion#3) #18660
- Linux capture module (PipeWire per-node capture): Verify Linux PipeWire capture on a real call and publish Linux release assets (code merged in quorum-companion#4) #18661
- Optional: spoken replies through a virtual microphone
Owner steps (no secrets in issues or PRs)
- Nothing to buy for the companion.
- Optional, for stable macOS permissions across updates: create a self-signed code-signing certificate and add it to
hivecommons/quorum-companionas the Actions secretsMACOS_SIGNING_P12andMACOS_SIGNING_P12_PASSWORD. Until then, releases are ad-hoc signed.
Sources
- Zoom free plan features (dial-in is a paid upgrade, 40-minute limit): https://www.zoom.com/en/products/virtual-meetings/features/free-video-conferencing/
- Zoom Audio Conferencing (toll dial-in, add-ons): https://www.zoom.com/en/products/virtual-meetings/features/audio-conferencing/
- Zoom Community: dial-in requires a licensed host: https://community.zoom.com/t5/Meetings/Zoom-Dial-In-says-quot-This-Meeting-Does-Not-Allow-Phone-Dial-In/m-p/34368 and https://community.zoom.com/phone-4/generate-call-in-phone-numbers-for-free-account-81809
- Joining a Zoom meeting by phone (DTMF order): https://support.zoom.us/hc/en-us/articles/201362663
- Zoom Conference Room Connector (H.323/SIP, paid add-on): https://explore.zoom.us/en/products/zoom-rooms/room-connector/ and https://library.zoom.com/advanced-enterprise-services/conference-room-connector-explainer/web-controls-and-dialing-procedures
- Zoom Meeting SDK bot policy and the March 2, 2026 authorization change: https://developers.zoom.us/docs/meeting-sdk/linux/ , https://devforum.zoom.us/t/question-on-linux-sdk-usage-policy/141545 , https://developers.zoom.us/docs/rtms/
- Zoom Terms of Service (section 7, recording-consent duty): https://www.zoom.com/en/trust/terms/
- Zoom custom live streaming (Pro or higher, admin setting): https://support.zoom.us/hc/en-us/articles/115001777826-Live-streaming-meetings-or-webinars-using-a-custom-service and https://it.umn.edu/services-technologies/how-tos/zoom-livestream-youtube-or-custom
- Twilio US voice pricing (dial-in estimate): https://www.twilio.com/en-us/voice/pricing/us
- Plaud Desktop (bot-free meeting capture with meeting detection): https://global.plaud.ai/pages/plaud-desktop and https://support.plaud.ai/hc/en-us/articles/53793763082009-Meeting-detection
- Apple ScreenCaptureKit (
SCStreamConfiguration,captureMicrophone): https://developer.apple.com/documentation/screencapturekit/scstreamconfiguration and https://developer.apple.com/documentation/screencapturekit/capturing-screen-content-in-macos - Apple Core Audio taps (macOS 14.2+): https://developer.apple.com/documentation/coreaudio/capturing-system-audio-with-core-audio-taps
- Ad-hoc signatures and lost permissions after updates (Apple DTS): https://developer.apple.com/forums/thread/795739
- Homebrew 5.0.0 (
--no-quarantineremoved, Gatekeeper checks): https://brew.sh/2025/11/12/homebrew-5.0.0
Activity
hivecommons-hive commentedon Nov 17, 2026
🔒 Claimed by agent
scanner(agent on hive hosted-available-oke-11-placeholder-r05x). The claim lasts 2h0m0s without activity and renews while the work continues. Reply/unclaimto release, or/claimto take it over if you outrank the holder.Claim history
Karibhou(contributor on hive hosted-available-oke-11-placeholder-r05x)Karibhou(contributor; covering PR merged or closed)scanner(agent on hive hosted-available-oke-11-placeholder-r05x)feat(companion): desktop companion audio platform; release 0.3.0 (#32)
hivecommons-hive commentedon Nov 17, 2026
PR #18602 covers this issue: #18602
The PR body says this is partial/non-closing work (
Refs/Part ofor leave-open wording), so Hive will not add a closing keyword. GitHub will not show its issue-list PR icon without a closing keyword; the stage comment is the link.✨ macOS v1: call detection, consent prompt, capture and streaming to …
✨ cask: quorum-companion 0.1.0 (macOS desktop companion for Betty) (#336
✨ meetings: desktop companion meetings (/api/companion/*)
📖 docs: Betty on Zoom without a Zoom app, options and recommendation (#…
✨ meetings: desktop companion meetings (/api/companion/*) (#18611)
9 remaining items
hivecommons-hive commentedon Nov 17, 2026
Scanner decomposition: no hive PR opened this pass, because the hive-side work is all merged. The design record is #18602, the
companionplatform,/api/companion/*routes and consent attestation recording (attested_by,attested_at,consent_version) are #18611, the quorum v0.3.0 pin is #18622, and the per-user SHA-256 companion token with hub passthrough is #18643. What is left under "Next" is either a manual check or work inhivecommons/quorum-companion, so it is filed as focused sub-issues titled[scanner] ..., each withPart of #18596:The optional spoken replies through a virtual microphone stay on this issue. That work needs an owner decision and a Quorum
saypath forcompanionfirst, which isplatform_unavailablein v1.🐝 Hive Agent:
scanner| Instance:hosted-available-oke-11-placeholder-r05x| SHA:unknown— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.326
[-]📖 Betty on Zoom without a Zoom app: dial-in and audio-bridge options[/-][+]📖 Betty on Zoom without a Zoom app: desktop companion (chosen path) and options considered[/+]🌱 chore: top up v6 from v5 (2026-11-17, second pass) (#18689)
🌱 chore: top up v6 from v5 (2026-11-18) (#18929)
hivecommons-hive commentedon Nov 18, 2026
All children delivered: #18659, #18660, #18661.
Every tracked child of this issue is closed, so the hive is closing it as completed. Reopen it if something it asked for is still missing.
clubanderson commentedon Nov 18, 2026
This issue was closed automatically with the note "All children delivered". To be precise about what that covers:
Done and merged: the design record (#18602), the companion meeting platform and routes (#18611), the Quorum pin (#18622), the per-user companion token (#18643), the Quorum companion platform (quorum#32, released as v0.3.0), the macOS app (quorum-companion#2, released as v0.1.0) and the Homebrew cask (homebrew-hive#336).
Windows and Linux: the capture code is merged (quorum-companion#3 and #4) and unsigned experimental builds are attached to quorum-companion v0.1.1. They have never been run on a real call, and verifying them is not planned until someone needs them (#18660 and #18661 were closed as not planned).
Not done: spoken replies through a virtual microphone. That needs an owner decision and a Quorum speaking path for companion meetings first, so it moved to its own issue (#19091).
A live test of the macOS app on a real call is not recorded on this issue or its children.