ZeroMCP v0.1.1 — ~27% faster with schema caching. See what changed →
Python

2.0x faster throughput.
Production HTTP transport.

ZeroMCP vs zeromcp-mrexodia — HTTP benchmarks.

Throughput
3,528 req/s ZeroMCP
vs.
1,804 req/s zeromcp-mrexodia
Latency (p50)
0.27 ms ZeroMCP
vs.
0.45 ms zeromcp-mrexodia
Memory
30 MB ZeroMCP
vs.
13 MB zeromcp-mrexodia

HTTP Performance — Head to Head

Same hello tool. Same methodology. 5-minute sustained load in Docker. Starlette for ZeroMCP, built-in transport for zeromcp-mrexodia.

req/s p50 p99 Memory CPU
ZeroMCP (Starlette) 3,528 0.27ms 0.57ms 30 MB 0.28%
zeromcp-mrexodia 1,804 0.45ms 1.24ms 13 MB 0.04%

The tradeoff

zeromcp-mrexodia is a lightweight Python MCP server with a minimal built-in transport — it keeps things lean at 13 MB. ZeroMCP plugs into Starlette for production-grade async HTTP, which is why it delivers 2.0x more throughput (3,528 vs 1,804 req/s) but uses more memory (30 MB vs 13 MB). Neither is wrong. mrexodia is simpler to keep small. ZeroMCP is simpler to keep fast under real load.

Drop a .py file. It's an MCP tool.