feat: Consolidate 8 scattered MCP repos into unified MCPC orchestrator#93
feat: Consolidate 8 scattered MCP repos into unified MCPC orchestrator#93
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
…elay Co-authored-by: groupthinking <154503486+groupthinking@users.noreply.github.com>
![vercel[bot]](https://avatars.githubusercontent.com/in/8329?s=60&v=4){kind=small}
| const results: ToolChainResult['steps'] = []; | ||
| const variables: Record<string, unknown> = {}; | ||
|
|
||
| let lastOutput: unknown = undefined; |
There was a problem hiding this comment.
Unresolved variable references in tool chain arguments silently pass through as literal strings instead of throwing an error
| // ── Tool manager ─────────────────────────────────────────────────────── | ||
| case 'list_tools': { | ||
| result = listTools(args['tag'] as string | undefined); | ||
| break; |
There was a problem hiding this comment.
Missing depth tracking in execute_tool_chain allows unbounded recursion and stack overflow
| # MCP tools/call returns {content: [{type, text}], isError?} | ||
| content = result.get("content", []) | ||
| if result.get("isError"): | ||
| text = content[0]["text"] if content else "Unknown error" |
There was a problem hiding this comment.
| text = content[0]["text"] if content else "Unknown error" | |
| text = content[0].get("text", "") if content else "Unknown error" |
KeyError raised when MCPC server returns error response with malformed content structure missing "text" key
| self._process.stdin.write((message + "\n").encode()) | ||
| await self._process.stdin.drain() | ||
|
|
||
| async def _read_loop(self) -> None: |
There was a problem hiding this comment.
Responses received for unknown request IDs are silently ignored, masking potential synchronization issues or server bugs
|
|
||
| case 'state_delete': { | ||
| const deleted = deleteState(args['key'] as string); | ||
| result = { key: args['key'], deleted }; |
There was a problem hiding this comment.
Tool execution errors are returned as successful JSON-RPC responses instead of proper error responses
|
|
||
| const id = request['id'] ?? null; | ||
| const method = request['method'] as string | undefined; | ||
| const params = (request['params'] ?? {}) as Record<string, unknown>; |
There was a problem hiding this comment.
Missing MCP-level parameter validation for tool arguments causes protocol-level errors to be returned as tool execution errors instead of invalid parameter errors
EventRelay had MCP functionality fragmented across 8 separate repos with no single canonical server. This adds
mcp-servers/mcpc/as the unified TypeScript MCP orchestrator and wires it into EventRelay's Python backend via a subprocess adapter.New:
mcp-servers/mcpc/— canonical MCPC serverTypeScript package (JSON-RPC 2.0/stdio, MCP
2024-11-05) consolidating:round_table— multi-agent consensuslist_tools,execute_tool_chain(variable binding, zero-glue chaining)state_set,state_get,state_list,state_deleteNew:
src/mcp/mcpc_adapter.pyAsync Python adapter that spawns the MCPC Node process over stdio and exposes a typed
call_toolAPI to the EventRelay backend:Registry + docs
src/mcp/mcp_registry.json—mcpcregistered ascanonical_orchestratormcp-servers/README.md— consolidation table added; MCP-management, MESH, mcp-tools-extension marked archivedOriginal prompt
💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.