Coverage join & session review markers¶
Two optional MCP workflows that most agents skip on the first pass:
- Coverage Join — attach an external Cobertura XML from your test run and preview untested-hotspot gating.
- Session review markers — track which findings you already triaged inside one long MCP process.
Start with Analyze & triage for health checks and PR review. Use this page when you already have a coverage artifact or a long finding backlog in the same chat session.
Normative tool shapes: session tools, analysis & gates, Coverage Join config.
Coverage Join (Cobertura + gates)¶
Join measured coverage to function hotspots for the current run only. Coverage Join does not update baseline, cache, or canonical report persistence.
| Requirement | Detail |
|---|---|
| Analysis mode | analysis_mode="full" — coverage_xml is rejected in clones_only |
| Input | Cobertura XML path on analyze_repository (coverage_xml) |
| Typical follow-up | get_report_section(section="metrics_detail", family="coverage_join") |
| Gate preview | evaluate_gates(fail_on_untested_hotspots=true, coverage_min=50) |
analyze_repository(root=<abs>, coverage_xml="coverage.xml")
-> get_report_section(section="metrics_detail", family="coverage_join")
-> evaluate_gates(fail_on_untested_hotspots=true, coverage_min=50)
CLI equivalent
codeclone --coverage coverage.xml --fail-on-untested-hotspots uses the same
join semantics. MCP evaluate_gates previews exit reasons without mutating
repository state.
Session review loop (in-memory markers)¶
MCP keeps run snapshots and session-local reviewed markers in the server process. They survive across tool calls but disappear on process restart — not Engineering Memory, not baseline truth.
Use when triaging many findings in one agent session: mark handled items, then filter them out on the next pass.
| Tool | Purpose |
|---|---|
mark_finding_reviewed |
Mark one finding reviewed (optional note) |
list_findings(exclude_reviewed=true) |
Omit findings already marked in this session |
list_reviewed_findings |
List markers for audit |
clear_session_runs |
Reset in-memory runs, markers, and workspace intent registry |
list_findings
-> get_finding -> mark_finding_reviewed
-> list_findings(exclude_reviewed=true) -> ...
-> clear_session_runs # full session reset — also clears active intents
For durable facts across sessions, use Memory recipes instead of review markers.