Show HN: MCP-C – cloud platform for running MCP agents and apps (docs.mcp-agent.com)

MCP-C (mcp-agent) is a unified platform and CLI for building, running and deploying MCP agents and MCP servers that scale from local development to a managed cloud runtime. It exposes a single MCP protocol surface (call_tool, read_resource, list_prompts) so the same code can be used with Claude, Cursor, ChatGPT Apps or custom SSE clients. The programming model relies on decorators like @app.tool, @app.async_tool and @app.workflow that run on asyncio locally and on Temporal in the cloud, giving you durable workflows with pause/resume, retries and human-in-the-loop support without rewriting logic. The project supports four deployment paths: managed mcp-c (mcp-agent deploy) where workflows run on hosted Temporal and stdio/ FastMCP servers run as sidecars; bring-your-own Temporal for air-gapped/on‑prem; plain MCP servers for stateless tools; and local iteration via mcp-agent dev start. Operational guardrails are provided by the CLI (build artifacts, secrets, auth, client install), observability streams logs through a cloud logger with OTLP export, and clients can connect over SSE/WebSocket using printed app hostnames and configured auth modes (bearer, OAuth, unauthenticated). For AI/ML teams this lowers friction to productionize agent-backed tools, reuse tool APIs across clients, and add reliability (Temporal) and traceability (OTEL) with minimal code changes.