Draft
Our service providers
Every company that may help us run NodNod, including our own in-house parts, what each one does, and what it may see.
Draft: pending legal review. This page is a working draft and may change before launch.
A service provider is a company that does one job for us, such as hosting the site, turning speech into text or taking a payment. Each one gets only what it needs to do that job.
Every part of a Nod (its face, voice, brain, ears, tools and room) can be swapped from one provider to another. So this page lists every provider we use or may use, not only the ones running today.
Providers may not use customer content to train their models under our accounts. [confirm: each provider’s terms for our account match this.]
Where an owner connects their own account with a provider (for example their own AI model key), that provider’s terms with the owner apply too.
We update this page when providers change.
Last updated 2026-11-07
Our own, in-house
NodNod (in-house) facesIn use
Faces, cartoon and photoreal models, run on our own rented computers (GPUs, and CPUs for some cartoon faces).
May see: Our agent’s voice, to move the face’s mouth. A cartoon face is drawn in your browser from mouth shapes we send.
Based in: United States
NodNod (in-house) voiceIn use
The default voice, an open voice model run on our own rented GPUs.
May see: The words our agent is about to say.
Based in: United States
NodNod (in-house) live timing voice and speech to speechMay be used
Open models that listen and speak in one step, run on our own rented GPUs. Not live yet.
May see: Live room audio, and the words our agent says.
Based in: United States
AudioBurnerMay be used
Our own speech to text service, from our family of products.
May see: Live room audio, and the transcript text it makes.
Based in: United States
MeetingBurnerMay be used
Our sister meeting product. A Nod can join a MeetingBurner room as a member, on the owner’s own MeetingBurner account.
May see: The Nod’s voice, face and name in the room. The Nod hears the people in that room.
Based in: United States
Hosting and compute
VercelIn use
Hosts the website and the API.
May see: Everything that passes through the site and the API: account details, Nod settings, transcript text, IP addresses.
Based in: United States
SupabaseIn use
Our database and sign in.
May see: Account details, Nod settings, transcripts, credit and billing records, audit records.
Based in: United States [confirm: data region]
ModalIn use
The rented computers (GPUs and CPUs) that run a Nod while it talks.
May see: Live room audio, transcript text, and the Nod’s voice and face as they are made.
Based in: United States
RunPodMay be used
Rented computers (GPUs), if we move a Nod’s work there.
May see: The same as Modal: live room audio, transcript text, and the Nod’s voice and face.
Based in: United States
LiveKitIn use
The live audio and video rooms a Nod talks in.
May see: The audio and video of everyone in the room, including the Nod.
Based in: United States
AI models (language models)
AnthropicIn use
The default model that writes a Nod’s answers.
May see: Transcript text, the Nod’s instructions and knowledge, and results from tools the owner ticked.
Based in: United States
OpenAIOnly when an owner chooses it
A model that writes a Nod’s answers, on the owner’s own OpenAI key.
May see: Transcript text, the Nod’s instructions and knowledge.
Based in: United States
xAIOnly when an owner chooses it
A model that writes a Nod’s answers, on the owner’s own xAI key. Not live yet: Grok voice + brain, which hears the room or the meeting and speaks, on our account or on the owner’s own xAI key.
May see: Transcript text, the Nod’s instructions and knowledge; with Grok voice + brain, the room’s audio too.
Based in: United States
GoogleMay be used
A model that writes a Nod’s answers.
May see: Transcript text, the Nod’s instructions and knowledge.
Based in: United States
Speech to text
AssemblyAIIn use
Turns what people say in a room into text, live.
May see: Live room audio, and the transcript text it makes.
Based in: United States
Recall.aiIn use
In an outside meeting, the meeting’s own live transcript with speaker names, used as a Nod’s ears.
May see: Meeting audio, and the transcript text it makes.
Based in: United States (our account runs in its US region)
The viewer’s own web browserMay be used
Turns what a viewer says into text inside their own browser, if an owner chooses browser captions. Not live yet.
May see: What the viewer says. The browser’s maker may process it under the viewer’s own browser terms. [confirm: which browsers send speech to their maker.]
Based in: Chosen by the viewer
Voices
The default voice is our own (see "Our own, in-house"). These are the outside voices a Nod may use.
GoogleOnly when an owner chooses it
Turns a Nod’s words into speech.
May see: The words our agent is about to say.
Based in: United States
OpenAIOnly when an owner chooses it
Turns a Nod’s words into speech, on the owner’s own OpenAI key.
May see: The words our agent is about to say.
Based in: United States
AnamOnly when an owner chooses it
Turns a Nod’s words into speech with an Anam face, on the owner’s own Anam account. Not live yet.
May see: The words our agent is about to say.
Based in: United Kingdom
Faces and video
Our own NodNod faces come first (see "Our own, in-house"). We hand our agent’s voice to an outside face provider line by line, and it sends back the moving face. An outside face provider receives only our agent’s voice, never the people in the meeting. It does not get the transcript or the Nod’s instructions, and a stock face needs nothing else. If an owner also picks that provider’s voice (see "Voices"), it gets the words our agent is about to say as well.
AnamOnly when an owner chooses it
A photoreal face, on NodNod’s account or the owner’s own Anam account.
May see: Our agent’s voice. If an owner makes a custom face with them, what the owner gives them for it.
Based in: United Kingdom
SimliMay be used
A photoreal face.
May see: Our agent’s voice. If an owner makes a custom face with them, the photo the owner gives them for it.
Based in: Norway
HeyGen LiveAvatarMay be used
A photoreal face.
May see: Our agent’s voice. If an owner makes a custom face with them, the video the owner gives them for it.
Based in: United States
Meeting platforms
Recall.aiIn use
Joins a Zoom, Microsoft Teams, Google Meet or Webex meeting for a Nod. Its meeting bot runs on a larger machine that Recall.ai runs for us (4 processor cores), opens our stage page and carries the Nod’s voice and face into the meeting.
May see: Meeting audio, who joined and left, the meeting’s transcript with speaker names, and the Nod’s voice and face from our stage page. Meeting media is kept at most 1 hour and is deleted when the meeting ends.
Based in: United States (our account runs in its US region)
Zoom, Microsoft Teams, Google Meet and WebexOnly when an owner chooses it
The meeting services an owner sends a Nod into. The meeting’s host runs them, under the host’s own terms.
May see: The Nod’s name, voice and face, like any other member of the meeting.
Based in: United States
Tools an owner connects
Tool servers the owner connects (MCP)Only when an owner chooses it
Tools a Nod may use, only the ones the owner ticked.
May see: What the Nod sends when it uses a tool.
Based in: Chosen by the owner
SunoOnly when an owner chooses it
Writes a song for the Sing ability, on the owner’s own Suno credits.
May see: What the song should be about.
Based in: United States
Sign in, email and payments
GoogleMay be used
Sign in with Google, if an owner chooses it instead of an email code.
May see: The owner’s name and email address from their Google account.
Based in: United States
ResendIn use
Sends our emails: sign in codes, confirmations and notices.
May see: The owner’s email address and the email we send.
Based in: United States
StripeIn use
Takes payments for credit.
May see: Card details, which go to Stripe and never to us, and the payment amount.
Based in: United States
Analytics and error reports
No analytics or error tracking service runs on NodNod today. If we add one, it is listed here first.
Inside a MeetingBurner room
When a Nod joins a MeetingBurner room, MeetingBurner’s own providers handle that room under MeetingBurner’s terms. They are: LiveKit (rooms), AssemblyAI and AudioBurner (speech to text), Anthropic, OpenAI, xAI and Google (AI models), Anam (faces), Modal (computers), Recall.ai (outside meetings), Supabase (database), Vercel (hosting), Resend (email), Stripe (payments), Sentry (error reports, with private text removed) and PostHog (product analytics). [confirm: link MeetingBurner’s own provider list.]