Sub-processors

Frontier is a real-time AI sales-call coaching platform designed to enhance sales performance. This page details the third-party services, known as sub-processors, that Frontier relies on. It distinguishes between services that handle customer or call data (data-path sub-processors) and those used purely for internal development or operations (development tooling).

Data-Path Sub-processors

The following services process or interact with customer and call-related data within the Frontier platform.

| Sub-processor | What it processes | Transfer Basis / DPA / Notes
| Recall.ai | Provides the desktop capture SDK (@recallai/desktop-sdk) for audio/video capture, and calendar/meeting OAuth via Google, Microsoft, Zoom. Handles meeting audio/transcript ingestion. | :::caution[GAP — founder supplies] Transfer basis and DPA. Note: The cloud bot integration is deprecated; the Desktop SDK is the live capture path. | | Deepgram | Speech-to-text (STT) for the direct-audio path (model nova-3). Processes raw call audio directly streamed from the client. | :::caution[GAP — founder supplies] Transfer basis, DPA, and zero-retention/no-training commitments. | | Vercel | Hosts the Frontier Dashboard (Next.js) and documentation sites. Processes web requests, user session management for the Dashboard, and may receive product telemetry. | | Vercel | Hosts the Frontier Dashboard and documentation sites. It handles content delivery, serving the Next.js applications, and routing web requests. | | OpenRouter | An LLM gateway and router that routes requests to various LLM providers. It also serves as a fallback embedding provider. It is responsible for routing and may process call-related textual context. | :::caution[GAP — founder supplies] Transfer basis and DPA. | | Perplexity | Configured but not actively verified to process customer data. May be a candidate for web-search/research functionality. | | **** | Data-Path Sub-processors (continued) | | | **** | Data-Path Sub-processors (continued) | ** | | Cohere | Used for paraphrase/semantic-similarity detection over script/manifest content in the Dashboard. | | Vercel | Provides infrastructure for the Frontier Dashboard and documentation sites. Handles content delivery, serving Next.js applications, and routing web requests. | :::caution[GAP — founder supplies] Transfer basis and DPA. | | **** | Data-Path Sub-processors (continued) | | | | | **** | Data-Path Sub-processors (continued) | | | | | Vercel | Provides hosting services for the Frontier Dashboard and documentation sites. Handles content delivery, serving Next.js applications, and managing web requests. | | OpenAI (via AI SDK) | Processes parts of the call transcript and knowledge base content for advanced answer generation, question detection, and query rewriting, utilizing various GPT models. | | Cloudflare | Provides core computing, networking, and data storage services including Workers (hosting the Call Server and its auxiliary services), **D ** (transactional database for live transcripts), Vectorize (vector index for embeddings/retrieval), R2 (object storage for AI Search source documents and crawled site storage), Workers AI (AI model inference), and AI Search (knowledge base backend for retrieval). | :::caution[GAP — founder supplies] Transfer basis and DPA for each specific service. Data residency details. | | Supabase | Processes application and account data. Stores org/user records (synced from Clerk), call metadata, participants (PII: names/emails), FAQs, scripts, org facts, AI answers, and calendar data. Also stores original uploaded knowledge files in Supabase Storage. | :::caution[GAP — founder supplies] Transfer basis and DPA. Data retention periods and data residency information. | | Perplexity | Processes potential web-search queries for various research functionalities. | | N | Cloudflare | | | OpenAI (via AI SDK) | Processes parts of the call transcript and knowledge base content for advanced answer generation, question detection, and query rewriting, utilizing various GPT models. | | Amazon Bedrock | Processes call transcript/KB content for answer generation, question detection, query rewriting, utilizing various Claude models, and potentially other AI models. It is a secondary LLM provider. | :::caution[GAP — founder supplies] Transfer basis and DPA, zero-retention/no-training commitments.