Subprocessors
1. Core infrastructure
These vendors process data as part of running the platform for every workspace.
| Subprocessor | Purpose | Data it touches | Location |
|---|---|---|---|
| Vercel | Hosting, CDN/edge network, serverless compute, scheduled jobs, page cache; Speed Insights performance monitoring on published pages | All request traffic, including visitor IP addresses (transient); geolocation derived from Vercel’s own request headers — we perform no third-party IP lookups; page-load performance measurements (Core Web Vitals, page URL, device and browser type) from visitors to published pages, without cookies | United States |
| Supabase | Postgres database and file storage for customer page assets | Account data (users, teams), funnel and page content, pseudonymous analytics events, encrypted integration tokens | United States (us-west-2) |
| WorkOS | Authentication (AuthKit) | User emails, names, and sessions | United States |
| Upstash | Redis — rate limiting and visitor-salt infrastructure | Hashed and transient identifiers only | United States |
| PostHog (US cloud) | Marketing website and product analytics for granvl’s visitors and users (not customers’ page visitors); dashboard session replay with text inputs masked; error tracking | Marketing page visits, traffic sources, CTA clicks and pseudonymous visitor IDs; platform-user events (team and user IDs, feature usage); dashboard session recordings | United States |
| Better Stack | Server logs and uptime monitoring | Application logs, which may include user and workspace identifiers | United States |
| OpenRouter | Internal LLM used for funnel analysis, daily briefs, and insights | Aggregate analytics text (no visitor PII by design); its downstream model providers act as sub-subprocessors | United States |
| Exa | Web search for research features | Search queries and the public web pages returned for them; no lead contact details | United States |
| Firecrawl | Fetching public web pages for research, brand-kit extraction, and page-style cloning | URLs you or your agent provide and the public page content retrieved; no lead contact details | United States |
| Resend | Transactional email (invites, weekly digests, funnel alerts) | Workspace-member email addresses and aggregate workspace statistics | United States |
| GitHub (Microsoft) | Source hosting, CI, and encrypted nightly database-backup artifacts (30-day retention) | Production database backups | United States |
| Amazon Web Services, Inc. (S3) | Storage of encrypted lead receipts | Browser-encrypted copies of form submissions; granvl holds no key able to read them | Canada (ca-central-1) or United States (us-east-1), per the customer's workspace setting |
2. Marketing capabilities — model and data providers
Used only when a workspace uses a Capability. “What we send” is the complete list for each provider; nothing else leaves the platform.
| Provider | Role | What we send | Served from | Retention / training | Terms |
|---|---|---|---|---|---|
| OpenRouter, Inc. | Routing layer for every generation model below | The prompt, chosen reference images, generation settings, an opaque per-workspace id | United States | Prompt logging off; we require providers that do not train on requests | Terms |
| OpenAI (GPT Image 2) via OpenRouter | Image generation | As above | United States | API data not used for training | Terms |
| Google (Nano Banana 2.1 via the paid Gemini API; Veo 3.1 via Vertex AI) via OpenRouter | Image and video generation | As above | United States | Paid API data not used for training | Terms |
| Black Forest Labs (FLUX.2 Pro) via OpenRouter | Image generation | As above | European Union | No training on API requests | Terms |
| ByteDance / BytePlus (Seedream 4.5, Seedance 2.5) via OpenRouter | Image and video generation | As above | Singapore | Per BytePlus AI terms; opt-out of training requested through the routing layer | Terms |
| Perplexity (Sonar, Sonar Pro) via OpenRouter | Cited web research | The research question only | United States | Per Perplexity API terms | Terms |
| Parallel Web Systems | Deep customer and market research | The offer description, market, optional landing-page URL and competitor domains the user supplied | United States | Per Parallel terms | Terms |
| DataForSEO | Search keyword volumes, costs and competition (Google Ads data) | Seed keywords, the market, optional landing-page URL and competitor domains | European Union | Per DataForSEO terms | Terms |
| Foreplay | Advertising library (what ads brands are running) | A brand name, domain or search query | United States | Ad text we retrieve is kept for at most 30 days; media stays on Foreplay’s servers | Terms |
| Browserbase | Hosted browser that loads a landing page for review | The page URL | United States | Session ends after the page is read | Terms |
A workspace owner can turn off any region (United States, European Union, Singapore) in Settings → AI / MCP; models served from a region that is turned off are hidden and refused.
3. Customer-connected integrations
These platforms process data only when you connect or configure them for your workspace, and only for that workspace.
| Platform | When it applies | Data it touches | Location |
|---|---|---|---|
| Meta Platforms | Only when you connect a Meta ad account: ad-account and spend sync in; server-side Conversions API events out, using your own pixel and token | Ad entities, spend, and creative metadata; conversion events you direct us to send | Global |
| PageSpeed Insights API (public page URLs only); GA4 Measurement Protocol when you configure it | Page URLs; conversion events you direct us to send | Global | |
| TikTok | Events API, only when you configure it | Conversion events you direct us to send | Global |
4. What is not a subprocessor
granvl is designed so that the most sensitive data path — your visitors’ form submissions — does not pass through our servers at all:
- Form destinations (Formspree, Zapier, GoHighLevel, HubSpot, ActiveCampaign, LeadProsper, Forms.app, your own webhooks or database) are recipients you choose. Form data is sent from the visitor’s browser directly to your chosen destination — granvl servers are never in that data path, so these providers are your vendors, not our subprocessors.
- Shopify is an inbound, customer-configured source of order values — it does not process data for us.
- Page generation by your own AI agent (for example Claude) happens through your agent and your account with that provider; granvl does not proxy or process those interactions.
5. Changes to this list
We will update this page before engaging a new subprocessor or materially changing an existing one, and will notify workspace owners by email at least 30 days before the change takes effect for their data. If you have a reasonable, data-protection-related objection to a new subprocessor, contact us within that period and we will work with you in good faith; if we cannot offer an alternative, you may terminate the affected service and we will delete your data as described in our Privacy Policy.
6. Contact
Questions about this list or our vendors: support@granvl.com.
granvl · British Columbia, Canada