feat: add MCP 2026 support - #158
Conversation
commit: |
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #158 +/- ##
=======================================
Coverage ? 94.41%
=======================================
Files ? 35
Lines ? 5196
Branches ? 2026
=======================================
Hits ? 4906
Misses ? 238
Partials ? 52 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
ec9dda9 to
07f797e
Compare
07f797e to
12e340a
Compare
douglance
left a comment
There was a problem hiding this comment.
Thermo-nuclear maintainability review: blocking on decomposition.
This PR takes src/Mcp.ts from roughly 165 lines to roughly 1,450 lines, crossing the 1,000-line boundary in one change. The new file now owns protocol constants and types, stdio framing, HTTP dispatch, authorization, capability negotiation, tool calls, resources, prompts, completions, task storage/lifecycle, elicitation adapters, caching, and JSON-RPC errors. That is not one cohesive module; it is an MCP subsystem concentrated in a single file.
Please decompose before merge. A clean boundary is: protocol/types, transport/stdio, transport/http, dispatcher, tasks/store, resources-prompts, authorization, and JSON-RPC errors. Keep Mcp.ts as the small public facade that composes those pieces. Mcp.test.ts also grows to about 1,260 lines and should follow the same ownership split.
This is the explicit under-1k to over-1k presumptive blocker. The behavior can stay unchanged, but the subsystem needs real module boundaries first.
Summary
Testing