Privacy policy

frntdsk runs AI agents that answer a business’s customers on voice, web chat, SMS and email, follow the procedures that business wrote, and take the actions it connected. That means we handle two kinds of personal data: information about you as our customer, and information about the people your agent talks to. This policy explains both.

Who we are

frntdsk is the data controller for your account, billing and usage data. For the conversations your agent has, the knowledge you upload and the systems you connect, you are the controller and we act as your processor: we only use that data to run your agent the way you configured it.

If you are a customer of a business that uses frntdsk and you want to know what was kept from a call or a chat, contact that business first. We will help them answer you.

What we collect from you

When you create an account we collect your name, email, company name, industry and the phone numbers you give us: your own, and the number you assign to your agent. Your password is stored as a bcrypt hash and your API key is stored as a hash, so neither can be read back by us. Sign-in issues a short-lived access token and a refresh token that lasts 15 days.

When you configure your agent we store what you author: procedures and their published versions, actions and the endpoints, headers and credentials they call, guidance, variables, the documents you upload to the knowledge base (file name, type, size, and the text chunks and embeddings made from it), your Composer threads and attachments, and the channels you connect.

Connecting a channel stores its identifier and settings: the phone number we provision for you, the domains your web chat widget may run on, the address you forward email from, or the Gmail account you authorise. A Gmail connection grants us read, send and modify access to that mailbox through Google’s API. We keep the connection tokens in our cache for 30 days from the last use, and you can revoke the connection at any time from the dashboard or your Google account.

For billing we keep your Stripe customer and subscription identifiers and the plan you are on. Card details are entered on Stripe and never reach our servers. We count conversations against your monthly quota and record overage when your plan allows it.

What we process on your behalf

Voice. Calls reach your agent through Twilio. The audio is streamed live to a speech to text provider and the words come back as a transcript; we do not record or store the call audio. We keep the transcript, the caller’s phone number, the agent’s replies, which were spoken through a text to speech provider, and the actions taken during the call.

Web chat and SMS. We keep the messages, the widget token and the origin domain of a chat, the sender’s phone number for SMS, and the requester’s IP address for rate limiting and abuse protection. Senders we block for spam stay on a blocklist for 30 days.

Email. For a connected mailbox or a forwarded address we read the messages your agent handles, draft replies, hold them for your approval when you ask us to, and send the ones you approve. Replies and their delivery records are kept with the conversation.

During any conversation your agent may collect the details your procedures ask for, such as a name, an address, an order number or a reservation, and pass them to the systems you connected. We store those details with the conversation as the variables your procedure defined.

About each turn we keep a trace: the intent we detected, the capability that ran, the procedure step, timing, a short reasoning summary and the result of any action. A model later reviews finished conversations and writes a recap, an outcome and findings so you can improve your agent. Reviews are for you; we do not use them for anything else.

How we use it

We use this data to run the service: to answer your customers, to run the procedures you authored and only those, to call the systems you connected, and to show you what happened in the dashboard. We use traces, reviews and recommendations to help you improve your knowledge base and procedures. We use account and usage data for billing, quotas, security, rate limiting and support. We do not sell personal data and we do not use your conversations to train models of our own.

AI processing and sub-processors

Answering a customer means sending the relevant parts of the conversation to the providers below. Each receives only what its job needs, under business terms with us.

• Groq and Anthropic run the language models that understand the conversation and write replies, review conversations and help you author procedures.
• Cohere turns your knowledge base and customer questions into embeddings and ranks results. Chroma, hosted inside our own infrastructure, stores those embeddings.
• Deepgram converts call audio to text. ElevenLabs converts your agent’s replies to speech.
• Twilio carries calls and SMS and provisions phone numbers. Postmark delivers and receives email. Google provides the Gmail API when you connect a mailbox.
• Stripe handles billing. Fly.io hosts the application and database in the United States. Upstash provides our Redis cache. Sentry and Axiom receive error reports and logs, which can include request details, so we can keep the service running.

We will tell account owners by email before adding a sub-processor that receives conversation content.

Where your data lives

The application, its Postgres database and the vector store run on Fly.io in the United States, in the San Jose region. Our providers may process data in other regions under their own terms. If you need your data to stay in a specific region, talk to us before you connect a channel.

How long we keep it

Conversations, messages, traces and reviews are kept while your account is active. You can delete a conversation or a document from the dashboard at any time, and deleting your account removes your conversations, documents, capabilities, variables and channels with it. Short-lived data expires on its own: OAuth sign-in state after 10 minutes, widget authorisation lookups after 5 minutes, files Composer generates for you after 24 hours, Gmail connection tokens 30 days after last use, and spam blocklist entries after 30 days. Logs and error reports are kept by those providers for their standard retention periods. Backups roll off within a few weeks of a deletion.

Security

Traffic to the API is encrypted in transit. Passwords and API keys are hashed, access tokens expire after 15 minutes, and refresh tokens are tied to one account. Webhooks from Twilio are checked against Twilio’s signature, inbound email from Postmark is protected with HTTP authentication, and a web chat widget only works with its token from a domain you allowed. Requests are rate limited and size limited, and admin endpoints need a separate credential. No system is perfectly secure, so if you believe your account has been compromised, rotate your API key from the dashboard and tell us.

Your rights and choices

You can see and change your account details, documents, procedures and channels in the dashboard, export your conversations through the API, and delete conversations, documents or your whole account. If you are in a region with data protection rights, such as access, correction, deletion, portability or objection, you can exercise them by contacting us and we will respond within 30 days. People who talked to your agent should contact you, and we will help you respond.

Cookies

This website sets no tracking cookies. When you sign in, the dashboard keeps your access and refresh tokens in your browser’s local storage so you stay signed in; signing out clears them.

Children

frntdsk is a business service and is not directed at children under 16. We do not knowingly collect their data as a controller.

Changes to this policy

When we change this policy in a way that affects you, we email account owners before the change takes effect and update the date at the top of this page.

Contact

Questions about privacy, requests about your data, and reports of a security issue go through our contact page.