Claude Code MCP connection timeout: startup wait vs a Sume call
Claude Code 2.1.284 waits up to 10 s for a connecting MCP server on resumed calls and 2 s on a non-interactive first turn. Sume tool calls hold at most 55 s.

Claude Code's connection waits are short and apply only while a server is still connecting. As of 2.1.284 (September 28, 2026), a tool call in a resumed session waits up to 10 seconds for its server, and a non-interactive first turn waits up to 2 seconds for servers named by --allowedTools or an mcp_tool hook, even when CLAUDE_CODE_MCP_STARTUP_WAIT_MS is 0. Once Sume is connected, a different limit applies: each remote call holds at most 55 seconds.
Claude Code facts are from its changelog; Sume facts from the MCP quickstart, MCP OAuth and API keys and Jobs and results, read 2026-09-30.
Which waits did 2.1.284 add?
Two entries matter. A resumed-session fix: tool calls no longer fail with "No such tool available" while the server is still connecting; the call waits up to 10 seconds for it. And a headless one: the non-interactive first turn still waits up to 2 seconds for connecting servers named by --allowedTools or an mcp_tool hook, even when CLAUDE_CODE_MCP_STARTUP_WAIT_MS is 0. The snapshot I read does not document other connection-timeout variables, so this post does not describe any.
How does a Sume server connect in claude -p?
Sume's server is plain HTTP at https://mcp.sume.com/mcp, with either OAuth or an API key. In a headless run there is no browser, so the docs' API-key path is the fit: send Authorization: Bearer $SUME_API_KEY or the x-api-key header. The docs note that API-key sessions can see write and paid tools, and that paid calls still need an idempotency_key.
Is the tool-call limit the same thing?
No. The waits above end when the server is connected or the wait elapses. After that, a Sume call that waits on a job holds at most 55 seconds, defaulting to 50, and the docs say to retry jobs_wait with the same ids on wait_slice_expired and never resubmit the paid create.
| Limit | Applies to | Value |
|---|---|---|
| Resumed-session call | Server still connecting | Up to 10 s |
| Non-interactive first turn | Servers named by --allowedTools or an mcp_tool hook | Up to 2 s, even with CLAUDE_CODE_MCP_STARTUP_WAIT_MS at 0 |
jobs_wait slice | One remote call to Sume | Default 50 s, max 55 s |
How do I check the connection first?
The quickstart's first call is tools_list or mcp_health. In a scripted run, make that your first allowed call so a connection problem shows up before any paid request. For the tool-call side see Claude Code MCP tool timeout with Sume video jobs.
Sources
Related posts
More in Developers
- Claude Code MCP error showed a Bearer token: rotate the Sume key
If an MCP error or log exposed a Sume Bearer value, rotate the API key. An OAuth token is not an API key. Report issues with the request id, not the key.
- Claude Code list_changed and reconnect: refresh Sume tools
Claude Code refreshes an MCP server's tools on list_changed and lets you reconnect from /mcp. After changing Sume's Write consent, verify with tools_list.
- Claude Code reserved MCP server name: widgets, and Sume's name
Claude Code 2.1.285 reserves the MCP server name widgets in cloud sessions and on self-hosted runners. Sume's documented name is sume, so nothing to rename.
- claude -p --permission-prompt-tool with Sume paid jobs
Headless claude -p can route permission prompts to an MCP tool. For Sume paid jobs, keep idempotency_key stable and wait with jobs_wait in 55s slices.
Written by Sume