Key concepts

TraceX uses a small set of terms consistently across the dashboard, Slack messages, emails and these docs. If a page uses a word you don’t recognise, it is defined here.

Your workspace

Term	What it means
Organization	Your company’s TraceX workspace. It holds your team, plan, billing, security settings and every project. See Projects and organizations.
Project	One product or production system, with its own connected sources, investigations, conversations and knowledge base. Most teams start with one project per product.
Member	A person in your organization. Each member has a role: admin, investigator or viewer.
Seat	What lets a member make changes. Members without a seat can still view everything. See Seats.

Connecting your systems

Term	What it means
Integration	A connection to another tool, such as GitHub, PostgreSQL, Datadog or Slack, set up per project under Integrations. See Integrations overview.
Source	An integration TraceX reads evidence from: a repository, a database, a log store or an error tracker. Plans limit how many sources a project can have.
Intake	An integration that sends issues to TraceX, such as a Slack channel, a support inbox, Jira, Zendesk or PagerDuty. Intake starts investigations without anyone opening the dashboard.
Read-only	TraceX never writes to the systems it reads. Database queries are checked before they run and refused unless they only read. See How TraceX handles your data.

Investigations

Term	What it means
Issue	Something your team wants explained: an error, a failing page, a wrong number, a customer complaint, an alert.
Investigation	TraceX working one issue to its root cause. Each has an id like TX48211794554145, a live page in the dashboard and a report at the end. See Start an investigation.
Triage	The first step of an investigation: working out what kind of problem it is (data, code, a recent change, an outside service) and where to look first.
Evidence	What TraceX found and relied on: a query and its rows, log lines, an error and its stack trace, a file and line, a commit. Every claim in a report links to evidence.
Root cause	The underlying reason for the issue, as opposed to its symptom. “Orders stay in pending_capture” is the symptom; “the capture job skips orders whose amount is in dollars instead of cents” is the root cause.
Confidence	How sure TraceX is of the root cause, and what would make it surer. Low-confidence answers say so plainly.
Ruled out	The other explanations TraceX checked and why they don’t fit. It shows how far the investigation looked.
Suspect changes	The deploys, commits and pull requests that most likely caused an incident, ranked with a reason for each. See How TraceX works.
Fix	The change that resolves the root cause: a code patch, a data correction or a configuration change, with its risk and how to check it worked. See Fixes and pull requests.
Share link	A read-only link to a finished report for someone outside TraceX. It never includes raw data. See Share a report.

Asking questions

Term	What it means
Conversation	A chat with TraceX about one project, for questions rather than incidents: “how many signups yesterday?”, “where is the refund state set?”. See Conversations.
Knowledge base	The runbooks, architecture notes and postmortems you give TraceX so it knows your system the way your team does. See Knowledge base.
Company profile	What your team has written about its systems (what a table or module is for) plus the facts TraceX learned from your corrections.
MCP	The Model Context Protocol, the standard AI coding tools such as Claude Code and Cursor use to call outside tools. TraceX is an MCP server. See Claude, Cursor and MCP.

Safety

Term	What it means
Masking	Replacing secrets, and optionally personal data, with placeholders before any text is sent to an AI model. See Data controls.
AI safeguards	Checks on every question and every piece of content TraceX reads for prompt extraction, jailbreaks and hidden instructions, and on every answer before it is shown.
Data access log	The record of every query, log search and file read TraceX made, with the investigation or question it was for. See Audit and access logs.