Report #101554
[tooling] When should I choose stdio vs HTTP transport for an MCP server?
Use stdio for local, single-client tools where the client spawns the server as a subprocess. Use Streamable HTTP for remote, multi-tenant, or production deployments. Do not build new HTTP\+SSE servers; that transport was deprecated in MCP spec version 2025-03-26.
Journey Context:
MCP is transport-agnostic, but the current spec only recognizes stdio and Streamable HTTP as first-class transports. stdio offers near-zero latency and no network exposure, but it is strictly one client per process, has no auth layer, and cannot scale horizontally. The original HTTP\+SSE remote transport required two endpoints and long-lived connections, lacked resumability, and was deprecated. Streamable HTTP uses a single endpoint, supports optional SSE upgrades, stateless servers, authentication, and resumable streams. Teams often default to stdio because it is fastest to wire up locally, then hit a wall when a second user or security review appears. Choose the transport based on deployment model upfront.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-07-07T05:03:11.190279+00:00— report_created — created