Roo Code MCP timeout 60 s default: add Sume as streamable-http

In Roo Code, set type to streamable-http, point url at https://mcp.sume.com/mcp, and leave timeout at the 60-second default. A Sume jobs_wait slice is 50 seconds by default and 55 at most, so it returns before Roo gives up. Roo accepts 1 to 3600 seconds, but a larger value does not lengthen Sume's slice.

Entry

Put it in mcp_settings.json for all projects, or in .roo/mcp.json for one repository. When both define a server with the same name, the project file wins, per the Roo docs.

{
  "mcpServers": {
    "sume": {
      "type": "streamable-http",
      "url": "https://mcp.sume.com/mcp",
      "timeout": 60
    }
  }
}

Choosing a timeout

Tool calls other than waits return fast. The long ones are jobs_wait and script_run, whose timeout_seconds goes from 5 to 55. Sixty seconds covers both with five seconds to spare.

Auth

OAuth sign-in is the preferred interactive route and gives read-only tools unless Write is toggled at consent. For a fixed key, use headers with Authorization: Bearer, and keep .roo/mcp.json out of commits if it holds one.

Global versus project file

Use the global file for a personal setup and the project file when a team shares the server. Since the project definition wins on a name clash, name the Sume entry the same in both to let a repository override your global timeout or tool lists. Keep credentials out of the project file.

Checking the connection

After saving, open Roo's MCP panel and confirm the server shows as connected, then ask it to call tools_list. If it fails, try mcp_health, which reports the endpoint and the auth source, and check that the entry type is streamable-http, since another type selects another transport.

If you change the timeout, change it for a reason you can state. Lowering it below 55 seconds makes a normal jobs_wait slice fail on the client side, so the agent sees an error when the server was about to answer. Raising it above 60 buys nothing for waits, because the server stops at 55. The one case for a higher value is another server in the same file that needs it, which Roo lets you set per server.

Before you rely on this setup, run a short acceptance test with a read-only credential. Connect, call mcp_health, call tools_list, and read one job with jobs_status. Record the tool count you see, so you can notice later if a credential change alters it. Then repeat the test after any config edit. A five-minute test like this catches most wiring mistakes before they cost money, and it gives you a baseline to compare against when something behaves differently next week.