feat(mcp): add get_why to the lean profile; present get_answer as the entry point

[swati510]

Problem

Two salience gaps, both measured by running the lean surface under a real coding agent on a mixed question set:

1. The lean profile omitted get_why. On why/history questions — the one category a code-search surface cannot answer from the tree alone — the agent had nothing better than file reads and scored measurably below an unassisted baseline. The differentiating tool was configured out, not absent.
2. Agents routinely paid a search_codebase round-trip before get_answer, even though get_answer runs the same hybrid retrieval internally — the tool descriptions never said so.

Change

• get_why joins LEAN_TOOLS (~2.1k tokens of schema total, still comfortably always-loadable). Docs updated.
• get_answer describes itself as the single entry point for questions; search_codebase redirects question-shaped queries there and positions itself for raw-hit use (enumeration, symbol_id resolution, scoping get_context).

Measured effect

With get_why served, the agent used it unprompted on history questions and the why-category score moved from below the unassisted baseline (7.3) to clearly above it (8.8); 21 of 32 runs opened directly with get_answer.

Testing

test_tool_selection.py, test_mcp_docs_drift.py (the drift guard enforced the docs update) — 19 green.

[repowise-bot]

✅ Health: 7.6 (unchanged)

📋 At a glance
3 hotspots touched · 2 files with recent fix history · 1 dead-code finding · no tests touched. Scoped to packages.

🩹 Review priority _{(files here with the most recent bug-fix history — defects cluster, so review these first)}

• .../mcp_server/tool_search.py — fixed 8× in the last ~6 months
• .../tool_answer/answer.py — fixed 3× in the last ~6 months

🧪 No test files were touched in this PR — consider adding coverage for the files above.

🔎 More signals (2)

🔥 Hotspots touched (3)

• .../mcp_server/tool_search.py — 15 commits/90d, 11 dependents · primary owner: Raghav Chamadiya (82%)
• .../mcp_server/_tool_selection.py — 3 commits/90d, 6 dependents · primary owner: Raghav Chamadiya (100%)
• .../tool_answer/answer.py — 19 commits/90d, 2 dependents · primary owner: Swati Ahuja (51%)

💀 Dead code (1 finding)

• 💀 .../tool_answer/answer.py _json_default (confidence 0.65)

---

📊 Full report · ⭐ Star Repowise · 📥 Install bot · Last updated 2026-07-17 11:53 UTC
_{Silence on a single PR with [skip repowise] in the title · Per-repo toggle on repowise.dev/settings?tab=bot}