Codex MCP tool_timeout_sec defaults to 60: does Sume jobs_wait fit?

Codex reads MCP servers from config.toml. For a streamable HTTP server the keys include url, bearer_token_env_var, http_headers, env_http_headers and auth. startup_timeout_sec defaults to 10 and tool_timeout_sec to 60 (read 2026-10-03). Sume's jobs_wait holds a single call for at most 55 seconds, so the 60 second default leaves five seconds of slack. That is enough, but raise it if your network is slow, and always re-call jobs_wait rather than resubmitting a paid create.

Authenticate with an env var so the key stays out of the file: bearer_token_env_var = "SUME_API_KEY".

Config

[mcp_servers.sume]
url = "https://mcp.sume.com/mcp"
bearer_token_env_var = "SUME_API_KEY"
startup_timeout_sec = 20
tool_timeout_sec = 90
enabled_tools = ["mcp_health", "tools_list", "tools_schema", "jobs_wait"]

Numbers side by side

Limit what Codex can call

enabled_tools is an allowlist and disabled_tools is applied after it; default_tools_approval_mode accepts auto, prompt, writes and approve. Pair a read-only allowlist with approval for writes if the agent should not start paid generation unattended.