Skip to content
Sign inRequest access

Subprocessors

uMarker uses a short list of third parties. The governing version is the legal list at umarker.com/en/subprocessors — Annex 3 of the Data Processing Agreement. This page explains it in plain words; if the two ever differ, the legal list wins.

Party Purpose What it can process Region
Microsoft Azure The servers for the API, the database, both portals and the public sites All service data India (Central India)
Backblaze B2 Backups of the database and files A copy of all service data United States (US East)
Cloudflare DNS, proxying and protecting every request, incoming email routing, storage of screenshots and uploaded files (R2), the bot check on sign-in pages (Turnstile), cookieless visit counts on the sites Requests in transit, IP addresses, incoming email, screenshots and uploaded files Global network; files stored in Western Europe
Google Analytics Visit measurement on umarker.com and these docs only, after the visitor accepts “Analytics” Data about visitors to those two sites only. Nothing from the dashboard, no customer data United States and global (per Google)
Microsoft Clarity Heatmaps and session recordings on umarker.com and these docs only, after the visitor accepts “Analytics”. All page text and everything typed into fields is hidden. Microsoft acts as an independent controller here, and may use the data for its own purposes, including advertising Interaction data about visitors to those two sites only (clicks, scrolling, mouse movement, page layout), approximate location from the IP address, device and browser, and the _clck and _clsk cookies. Nothing from the dashboard, no customer data The Microsoft Azure cloud, the United States and other countries
Google (Gmail), via Cloudflare Email Routing The mailbox behind support@umarker.com Support messages and their contents United States and global
Groq AI triage suggestions, drafts and summaries — only if your account turns AI on Report text and support conversation text only, no screenshots United States
Google Gemini The same AI features, through the paid Gemini API, under which Google does not use the text to train its models — only if your account turns AI on Report text and support conversation text only, no screenshots United States and global (per Google)
Email providers Service email: invitations, verification, password resets, notifications, billing Recipient name and address, subject and body Varies by provider

Service email goes through an ordered chain of providers — a message leaves through the first one that accepts it, so any of them may carry any message. The legal list names each one with its company and processing locations.

Three things people usually ask about, stated as facts:

AI providers: on the list, off for your account

The platform’s AI chain today uses Groq and Google Gemini. Nothing goes to either for your account unless your plan includes AI — Professional and above — and an admin of your account switches AI consent on; it is off on every account by default. An account may instead connect its own key, and then its text goes to the provider it chose, under its own contract.

No payment processor

Billing is a manual bank transfer approved by a person. There is no card gateway, which means no third party, and no part of uMarker, ever handles card details, because none are collected.

No analytics or session replay in the dashboard

The dashboard ships no third-party analytics, tag manager or session recorder, so nothing about your team’s use of the product is streamed anywhere. Only the two public sites (umarker.com and these docs) use Google Analytics and Microsoft Clarity (heatmaps and session recordings with the text hidden), and only after a visitor accepts “Analytics” in the cookie banner; neither sees any customer data. Microsoft acts as an independent controller for Clarity; the cookie policy has the details.

What happens when your account turns AI on

Section titled “What happens when your account turns AI on”

The features exist and the platform has providers, so the honest thing is to describe what they send and what they do not, rather than leaving a gap for someone to guess at.

Turning them on takes three things: your plan includes AI — Professional and above — and the deployment’s operator has at least one model provider in its chain (on uMarker today: Groq and Google Gemini), and your account switches consent on. Consent is off on every account by default, because the reports carry your customers’ words and screens and you need to be able to tell those people what happens to them before anything happens at all. On Professional and above, an account may instead bring its own verified key; then its report text goes to the provider that account chose, under that account’s own contract, and the operator’s choice of provider no longer applies to it. The plan and the consent still do.

What is sent is bounded structurally, not by policy. The provider contract receives the report’s text: the comment, the console lines (scrubbed for credential shapes first), and the page URL. It has no path to the screenshot, the reporter’s email address, or the technical enrichment. Nothing is written automatically either; a suggestion lands in separate columns and a person accepts it.

The providers the code can be pointed at, with the host each one would receive text at:

Provider Host
Anthropic api.anthropic.com
OpenAI api.openai.com
Groq api.groq.com
Google Gemini generativelanguage.googleapis.com
DeepSeek api.deepseek.com
xAI api.x.ai
Mistral api.mistral.ai
Any OpenAI-compatible endpoint whatever the operator configures: a third party, or a model the operator hosts themselves
  • Redaction happens in the reporter’s browser before a report leaves the page, so whatever it masks never reaches uMarker, let alone anyone downstream. How much it masks is the project’s own setting: only the query-string blanking runs unconditionally. See PII redaction.
  • Media never goes to a model. Screenshots stay in the storage uMarker operates; the AI contract cannot reach them.
  • Your own integrations are your choice, not ours. If you route reports to Jira, Linear, GitHub, Slack, Teams, Telegram or your own webhook, those receive the report content you asked them to receive. They are your processors, under your contracts, and they do not appear in the table above.