MCP Verify Round 2 implementation notes

Date: 2026-08-04

This documents the implementation of the Round 2 public-site corrections for MCP Verify.

Implemented

• TASK-26: Reconciled Strong live candidates with a concrete public filter.
• Definition is now: public, non-hidden servers with status=healthy, score >= 70, tool_count >= 1, and validation within the last 48 hours.
• The stat links to the exact filter URL:

/?freshness=strong_live&min_score=70&status=healthy&tools_min=1&sort_by=score&sort_order=desc

• Added repository support for freshness=strong_live.

• TASK-27: Fixed default homepage ranking semantics.
• Default homepage ordering is now freshness tier first, then score descending, then last validated timestamp.
• The sort label now says Sort by freshness tier, then score.

• TASK-28: Made the default result-set constraint explicit.
• The homepage now shows a removable chip:

Default view: validated, healthy/degraded, score ≥60, tools ≥1

• Removing it navigates to /?all_servers=1, which expands toward the full public corpus.
• The count copy now states the default subset against the indexed-server total.

• TASK-29: Locked zero-tool exclusion.
• Zero-tool servers continue to resolve to metadata_only.
• Zero-tool servers are excluded from default homepage results and the strong-live metric/filter.

• TASK-30: Removed redundant discovered/scored display when counts match.
• If discovered and scored counts are equal, the coverage row renders one Indexed servers card instead of separate duplicate funnel stages.
• If they diverge in the future, both stages are rendered again.

• TASK-13 dependency: Status text is now visible in result rows.
• The status column renders the status dot plus readable text.

Verification

Commands run:

PYTHONPATH=verify/src .venv/bin/pytest -o addopts='' verify/tests -q

Results:

• Full Verify test suite: 223 passed, 3 warnings.