Data Integrations

Plugs into how your team already works.

Six read-only integrations covering your engineering workflow from code review to incident response. Each integration adds a distinct signal layer to your competency graph — and the graph gets sharper as more sources connect. You don't need all six on day one. Most teams start with GitHub or GitLab plus Jira and see meaningful gaps within 48 hours.

Request Access

GitHub

Reads PR review comment metadata, reviewer-to-contributor assignment patterns, file path coverage by domain, and PR velocity by engineer. Surfaces reviewer vs. contributor skill asymmetry — the gap between who can approve code and who can write it confidently. Uses pull_requests and issues read scopes only. No repository content access.

GitLab

Reads merge request patterns, pipeline failure context, review coverage across service boundaries, and contribution distribution per domain. Identifies where a single engineer is the only reviewer capable of approving changes in a given service — a skill coverage risk that rarely appears in surveys.

Jira

Analyzes ticket assignment patterns, escalation chains, stale epics, and domain concentration over time. Reveals where skills are concentrated in one or two engineers and where they're entirely absent. Ticket reassignment loops are a particularly strong signal of undocumented skill gaps.

Confluence

Maps documentation authorship and editing history to identify knowledge ownership and codification gaps. Surfaces where institutional knowledge exists only in one person's head — and where documentation exists but no engineer has hands-on experience to match it.

Slack

Reads public channel message metadata to identify help-seeking patterns and informal knowledge routing. Maps which engineers get pinged repeatedly in the same domains — revealing informal expertise nodes and learning bottlenecks that don't appear in any formal org chart. No private message content is read.

PagerDuty

Correlates incident involvement records with skill domains in the competency graph. Engineers repeatedly paged for the same incident type signal a skill gap in the broader team — or a dangerous single point of expertise. This is often the most direct link between an unresolved gap and a production risk.

What data we read — and what we don't.

Engineering teams get cautious when someone asks to connect to GitHub. Fair. Here's exactly what Tunlai reads and what it explicitly does not access.

  • PR review comment metadata (author, reviewer, timestamp)
  • File path patterns in commits (domain signal, not content)
  • Ticket assignment and escalation chains
  • Incident involvement records
  • Documentation edit history (Confluence)

We do not read:

  • Source code content (lines of code, functions, logic)
  • Private messages or DMs in Slack
  • Salary or HR data
  • Individual performance ratings
  • Anything not covered by the authorized OAuth scopes