Skip to content

v1.3.0 — Hallmark Online MCP

Choose a tag to compare

@Wolfe-Jam Wolfe-Jam released this 15 May 02:01
· 128 commits to main since this release

The hallmark-online-MCP release. Identity restored, receipts measured, doctrine aligned, four origin credentials surfaced through the wire + page + .faf.

What changed

Lane B — Identity restored (22+ sites). MCP initialize handshake now declares name: 'grok-faf-mcp' (was 'claude-faf-mcp' due to incomplete fork). Class ClaudeFafMcpServerGrokFafMcpServer. Install instructions, debug strings, .faf output tags, SKILL.md resolution paths all corrected.

Lane A — Comparison framing scrubbed (6 sites). Removed x-grok-wins: true header. Cleared "#1 model on Earth" from /health.dedication, /info.description, HTML hero, and the grok_go_fast_af tool description. @ElonMusk tribute preserved (shortened).

Lane C — Server-card enriched. /.well-known/mcp/server-card.json now declares protocolVersion (2025-06-18), capabilities block, instructions with the canonical credentials, transport, plus serverInfo expanded with description / homepage / repository. Endpoint advertised in /info.endpoints.

Lane D — Real measurement. Added bench/score.bench.js — reproducible Mk4 WASM scoring bench (npm run bench:score). Measured: 137 µs/score · 7,279 ops/sec p50 (3 batches × 1,000 iterations). About-modal now carries this real number; receipt at bench/results-2026-05-14.json holds full hardware + runtime + per-batch stats.

.faf refreshed with four origin credentials: first MCP for Grok · first FAF MCP online · MCPaaS pattern origin · IANA .faf + .fafm. Re-scored 🏆 Trophy 100% (15/15 slots).

CLAUDE.md → GROK.md per the MD doctrine — each frontier AI gets its own MD; AGENTS.md remains as the separate vendor-neutral file.

Numbers

  • 169 / 169 tests passing across 6 suites
  • TSC clean
  • Stats hero: 0.2ms Avg Response (conservative undersell of measured 137µs)
  • Tarball: 197 files · 314.5 kB

Receipts

FAF defines. MD instructs. AI codes.