What is Streamable HTTP?
Streamable HTTP is one of two transport modes supported by the Model Context Protocol (MCP). It enables AI clients to connect to MCP servers over HTTP with support for streaming responses, making it suitable for remote server deployments accessible over the network.
Streamable HTTP vs stdio
- Streamable HTTP – For remote deployments. MCP servers run as network services, and AI clients connect over HTTP. Supports streaming for long-running operations.
- Standard IO (stdio) – For local deployments. MCP servers run as subprocesses, and AI clients communicate via standard input/output. Used in IDE extensions and local agent setups.
In Naftiko
Naftiko capabilities that expose MCP adapters support both transport modes from the same deployment. This means the same capability can serve remote AI platforms via streamable HTTP and local IDE copilots via stdio without separate configurations.
Why it matters
Different AI consumption patterns require different transport modes. Remote agent platforms need HTTP connectivity, while local IDE copilots need subprocess-based communication. Supporting both from a single capability spec eliminates the need for separate deployments per transport mode.
