Allow configuring MCP slow-connection warning threshold
- 主要言語
- Shell
- スター
- 11.2k
- フォーク
- 1.9k
- 平均マージ
- 14時間 16分
- マージ済み PR(30日)
- 6
説明
### Describe the feature or problem you'd like to solve
MCP servers that require authentication handshakes (e.g., OAuth, Entra ID) routinely take 5-10 seconds to connect. The CLI displays a warning for each server that exceeds the hardcoded 10-second threshold:
```
! MCP server 'server-a' is taking longer than expected to connect.
! MCP server 'server-b' is taking longer than expected to connect.
! MCP server 'server-c' is taking longer than expected to connect.
```
All servers connect successfully -- the warnings are purely cosmetic. With many MCP servers configured, startup is dominated by a wall of yellow warnings that adds noise without actionable information.
The `timeout` field in `mcp-config.json` controls the hard connection timeout (how long before giving up), but there is no way to control the warning threshold.
### Proposed solution
Add support for a `slowConnectionThresholdMs` field in `mcp-config.json`, either:
**Per-server** (preferred):
```json
{
"mcpServers": {
"my-server": {
"type": "stdio",
"command": "my-mcp-server",
"args": ["start"],
"timeout": 60000,
"slowConnectionThresholdMs": 30000
}
}
}
```
**Or globally:**
```json
{
"mcpServers": { "...": "..." },
"slowConnectionThresholdMs": 30000
}
```
The internal plumbing already supports this -- the slow-connection monitor class constructor accepts `slowConnectionThresholdMs` as an option, it's just not wired to any user-facing config.
### Example prompts or workflows
1. User configures multiple MCP servers that authenticate via OAuth or similar
2. Each server takes ~5-10s to complete auth + upstream connect
3. User sets `"slowConnectionThresholdMs": 30000` to suppress warnings for expected auth latency
4. Startup shows clean output with no false-positive warnings
### Additional context
- The slow-connection monitor class already accepts `slowConnectionThresholdMs` in its constructor options -- this is a config-wiring change, not a new feature
- The default 10s threshold is hardcoded as a constant
- This primarily affects users with authenticated MCP proxies (OAuth servers, enterprise auth flows)
- A single OAuth/auth handshake can take ~5s, so even one hop can approach the 10s threshold
コントリビューションガイド
調査の方向性
Start with the mcp-config.json schema and configuration path, then inspect the slow-connection monitor class and its constructor options. Wire the proposed slowConnectionThresholdMs setting to the existing monitor and verify that a configured threshold prevents warnings for connections below it while retaining the default behavior when unset.
索引モデルが issue の本文から書いたものです。
評価
- 技術スタック
- shell
- 領域
- cli
- issue の種類
- 機能追加
- 難易度
- 3/5
- 見積もり時間
- 1〜2日
- 活発さ
- 静か
- 明瞭さ
- おおむね明確
- 初心者へのやさしさ
- 68/100