feat(mcp): expose arch-metrics, memory, export-snapshot, import-snapshot as MCP tools — closes #121

[Wolfvin]

What

Fixes RED TEAM issue #121: MCP server was missing tools for 4 new CLI commands added in #107, #109, #110. AI agents using CodeLens via MCP (Claude Desktop, Cursor, etc.) could not invoke these features.

What changed

Added 4 tool definitions with full JSON Schema parameter specs to scripts/mcp_server.py _TOOL_DEFINITIONS:

Tool	Parameters
arch-metrics	workspace, format (json/markdown/ai/sarif/compact, default: ai), top (default: 50)
memory	workspace, action (write/read/list/delete), name, content
export-snapshot	workspace, output
import-snapshot	workspace, input, merge (default: False)

MCP tool count: 50 → 54. CLI command count remains 67 (some CLI commands like init, serve, watch are intentionally not exposed as MCP tools because they don't make sense in an MCP context).

Why explicit definitions (vs auto-generation)

The MCP server's _dynamic_tool_definitions() mechanism (line 1387) already auto-generates tool definitions for any CLI command not in _TOOL_DEFINITIONS, but the auto-generated specs lack parameter schemas (everything is treated as opaque workspace + args). The explicit definitions added here give AI agents proper parameter discovery so they can call the tools correctly without trial-and-error.

Verification

>>> from scripts.mcp_server import _TOOL_DEFINITIONS as T
>>> len(T)
54
>>> all(k in T for k in ['arch-metrics', 'memory', 'export-snapshot', 'import-snapshot'])
True
>>> T['memory']['parameters']['properties']['action']['enum']
['write', 'read', 'list', 'delete']

Test suite: 1183 passed, 1 pre-existing failure (#118 — test expectation mismatch, unrelated).

Test plan

• 4 new tools appear in _TOOL_DEFINITIONS with correct parameter schemas
• Full test suite: 1183 passed (no new regressions)
• CI: full test suite passes
• Manual: MCP client (e.g. Claude Desktop) can invoke the new tools

Risk

Low. Only adds new entries to a dict — no existing tool definitions are modified. The MCP server's tool dispatch already handles these command names via the existing CLI command registry.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud