Skip to content
MCP ThesaurusMCP Thesaurus

Who Owns What

CommunityGood71/100Claim

MITupdated 7d ago

People are first-class nodes in the graph, with identities linked across vendors (GitHub, Jira, Slack, …) — one human = one identity. Ownership questions are graph questions.

SourceWebsiteDocs

What can you do with Who Owns What?


name: who-owns-what description: Find people — owners, experts, reviewers — through the org graph. Use when asked who owns, maintains, or knows a system, who to ask, assign, or request review from, who wrote or decided something, or when routing work to a human.

Who owns what

People are first-class nodes in the graph, with identities linked across vendors (GitHub, Jira, Slack, …) — one human = one identity. Ownership questions are graph questions.

Workflow

  1. Find the person or the artifact with bonez:search:
{"query": "payments service owner", "entities": ["user", "person"], "limit": 5}
  1. Traverse with bonez:query from either end — from an artifact toward its authors/reviewers/deciders, or from a person toward what they touch. The grammar and people-edge names live in the bonez:query tool description; check bonez:schema before guessing person-related edges.

  2. bonez:fetch the person record before naming them — don't guess emails or handles.

Judgment

  • Ownership is behavioral, not titular. Authority topology — who reviews, who merges, who gets asked — beats last-commit-wins. A recent PR author may be a drive-by contributor; the person who reviewed the last ten PRs in that directory is the owner-shaped signal.
  • Exclude bots. Bot accounts author enormous volumes of commits and comments. Never nominate a bot as an owner or expert; discount bot activity when weighing evidence.
  • Identity linking can be incomplete. Two similar-looking "people" (a GitHub login and a Jira account, say) may be one unlinked human. When the evidence splits oddly across near-duplicate identities, consider merging it mentally and say the identities may be unlinked.
  • People go stale faster than code. People leave; teams reorganize. Check the recency of the evidence before assigning — a two-years-ago expert may be gone. Prefer "most recent sustained activity" over "most total activity".
  • Small orgs, small graphs. With few indexed people, the top result may be the only candidate, not the best one. Report the strength of the evidence, not just the name.

Hand-offs

  • "Why was this decided" (rather than "who decided") → finding-prior-art.
  • Presenting a person with their evidence → citing-bonez-sources (vendor profile URL, not a bare handle).