Skip to content
Sign inRequest access

Core concepts & glossary

uMarker has a small vocabulary. Learn these dozen terms and the rest of the docs read easily. They fall into four groups: structure (who owns what), people (who does what), the report (what gets captured), and triage (how it moves).

Account: the tenant. A customer organization and the hard data-isolation boundary: every space, project, report, user and setting belongs to exactly one account, and one account can never read or write another’s data. Every account-owned record carries an account_id, and the boundary is held at each door rather than by one global switch: a tenancy scope on the account-owned models, explicit account filters on the paths that run before a principal exists, public-key project resolution on the anonymous surfaces, and a project-access check on anything addressed by id. Billing and the team live at this level, and so does a person’s role — which a single project can override.

Space: a grouping of projects, and the unit you navigate by. It is not optional and not a power feature: after you sign in you land on the spaces grid, and every project belongs to exactly one space. A space called General is created with your account, it can be renamed, given its own accent colour, and deleted like any other space.

Creating additional spaces is a paid capability, from Professional up: every account already has General, so the capability means “more than one”. On Free and Starter you work in a single space, which is the normal shape for a single product team.

Project: one site you collect feedback on, with a name, its domains (zero or more — a draft needs none, going live needs at least one), widget settings, distribution routes, a public key, and — from Professional up — targeting rules. It is the anchor everything else hangs off, and it is the thing you publish or leave as a draft.

Domain: a site mapping on a project (example.com). Domains do two jobs. They resolve an incoming page to at most one project, deterministically, so uMarker knows where a report belongs; and they are the allowed-domain list that decides where your widget is permitted to run at all. An empty list means the widget runs nowhere, and a project with no live domain cannot be published. Overlapping domains across projects in the same account are not allowed.

Published vs. draft: a project is published when its widget is live. Publishing mints a public key if the project does not have one and refuses if there is no live domain, so “published” can never mean “live but broken”. Moving a project back to draft stops the widget and keeps the key, so the snippet already pasted on your site does not break.

Team member: someone with dashboard credentials, invited by email. Two built-in seats ship: admin, which resolves to every permission, and member, which triages but does not administer. From Professional up you can define custom roles with a permission matrix, and hand someone a custom role on a single project, which overrides the role they hold on the account — so a client reviewer can see one project and nothing else. Every granting door carries a privilege-escalation ceiling: you cannot hand out a permission you do not hold yourself.

Extension tester: an identity for capturing, not triaging. These are the testers, clients and stakeholders who connect through the browser extension. A tester does not get a dashboard password: you add their email to a project and hand them a connection code (mdc_…), which is shown once. They are assigned to the projects you attach them to, and at any time you can issue a fresh code (which stops their live sessions), remove them from a project, or delete the tester outright.

Feedback (a report): the core object. One submission from the widget or the extension: a screenshot of what the reporter saw — one picture of the screen for each place they marked, taken as they marked it and stacked into one image — one or more numbered pins, each carrying its own comment, and the automatic context. What that context holds follows the project’s capture mode (Widget → Setup): Normal — what new projects start on — carries the page and device only; Technical adds console logs, network requests, errors, performance, reproduction breadcrumbs and environment. A report keeps the mode it was filed in. One further piece exists per project but ships off and has no control on any screen yet — the resolved element panel — so a report today carries it only if you set the flag over the API.

Mark (a pin): one annotation on a report. The capture tool produces marks of a few kinds: an element (what the inspector was pointed at, carrying its CSS selector), a margin (a click on the orange spacing band, carrying which side), and the free-hand kinds (comment pin, highlight and pen). Geometry is normalized 0–1 against the same area the screenshot covers, so marks replay exactly over the clean screenshot. In the dashboard every mark has its own comment thread and its own done state.

Redaction: the masking that happens in the reporter’s own browser before the screenshot is taken. Redaction is a property of capture, not a server-side afterthought. One part is always on: query-string values whose parameter names suggest a secret are blanked, on every project. Blurring visible field values and any region you name with a CSS selector is a per-project switch, off until you turn it on.

Capture pipeline: the path a report travels. Capture (browser) → ingest (server re-validates authorization, fail-closed) → triage (dashboard) → resolve → optionally close the loop. The same pipeline runs whichever surface captured it.

Status: where a report is in its life. The set is fixed: new, in_progress, waiting, resolved and archived. Every change is attributed to an actor and a timestamp in the audit log.

Priority: how urgent a report is. One of none, low, medium, high or urgent, set by a triager to order the board.

Assignment: the team member a report is handed to. One owner, accountable for moving it forward.

Label: a tag from the space’s shared taxonomy. The same label set applies to feedback and survey responses, so the signal shares one vocabulary. Your account ships with a four-label starter set — Bug, Idea, Question and Improvement — each carrying both an Arabic and an English name, so every reader sees their own.

Inbox: where reports land, as a board or a list. The board is drag-ordered and shows up to 100 cards per column with the true count beside it, so a column with twelve thousand reports says so rather than trying to draw them.

Public key: the md_… identifier the widget carries in its data-umarker-key attribute. It is public by design: it resolves one project and grants no access to your account on its own. Every real control is enforced on the server: the allowed-domain check, validation, rate limits, size caps, the published switch. Rotate or revoke it from the project’s Snippet screen; rotating resets the install status, because the old snippet is now dead.

API key: a secret credential for the REST API and the MCP server, stored hashed and scoped either to the whole account or to a single project — prefer the narrower one. Unlike a public key it authenticates programmatic access and must be kept private.

A few terms you will meet once you are past first capture:

  • Survey: a proactive NPS / CSAT / CES or form microsurvey shown through the same widget, with rating, stars, emoji, single-choice and text questions. Responses share the labels and export to CSV.
  • Announcement & changelog: in-app “what’s new” for your users, plus a hosted public changelog page you can link to.
  • Insights: a read-only rollup of the status split, a 14-day report trend, top labels across feedback and survey responses, and survey headline scores, per space or per project.
  • Distribution route: how a credential reaches a project. You connect a tracker or a channel once for the account, then route it per project with per-event switches, so an agency does not re-paste a Jira token for every client.