🤖 fix: add MCP server startup timeout and surface failures

[ibetitsmike]

Summary

Misconfigured MCP servers (e.g., wrong command+args format) cause startSingleServer() to hang indefinitely on createMCPClient() / client.tools(), blocking the entire chat stream. This adds a 60s startup timeout and surfaces failures to the user via a system message warning.

Background

• startSingleServer() awaits MCP client creation with no timeout — a broken server config hangs forever
• The existing runServerTest() (settings "Test" button) already has a Promise.race timeout, but the production startup path did not
• Errors were caught and logged but never surfaced to the user

Implementation

Startup timeout (mcpServerManager.ts):

• New MCP_STARTUP_TIMEOUT_MS = 60_000 constant
• Extracted startSingleServer → startSingleServerImpl, wrapped with Promise.race timeout following the existing runMCPToolWithDeadline pattern (cleanup via clearTimeout + .unref())
• Reduced stdio exec timeout from 24h to match startup timeout

Failure tracking (mcpServerManager.ts):

• startServers() now returns { instances, failedServerNames }
• MCPWorkspaceStats extended with failedServerNames: string[]
• Both call sites in getToolsForWorkspace() updated to propagate failures

User-facing warning (aiService.ts):

• When servers fail, a warning is prepended to the system message so the model informs the user which servers are unavailable — zero frontend changes needed

---

Generated with mux • Model: anthropic:claude-opus-4-6 • Thinking: xhigh • Cost: $4.32

[ibetitsmike]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 6ab14a2970

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[ibetitsmike]

@codex review

Addressed: updated the leased+closed restart path to merge failedServerNames into returned stats (commit df5be4e), so aiService.ts now correctly surfaces the warning.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: df5be4e517

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[ibetitsmike]

@codex review

Addressed: reverted the stdio exec timeout back to 24h (process lifetime), since startup timeout is already handled by the Promise.race wrapper in startSingleServer.

[ibetitsmike]

@codex review

The stdio exec timeout is already 24h (line 1367: timeout: 60 * 60 * 24). This was fixed in commit 7014420 — the previous thread was about the same issue and was already resolved. Startup timeout is handled separately by the Promise.race wrapper.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 70144207cd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[ibetitsmike]

@codex review

Addressed: added AbortController to startSingleServer so timed-out startups now abort the in-flight impl and clean up spawned processes. Both stdio and HTTP/SSE paths have abort listeners that close transports/clients on timeout.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e70d79d00b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[ibetitsmike]

@codex review

Addressed: token recount after MCP warning prepend. Uses getTokenizerForModel + resolveModelForMetadata to recount systemMessageTokens after mutating the system message.