LiteAPI CommandDEMO
DEMOIndependent demo built on LiteAPI's public sandbox and open-source MCP server. Not affiliated with Nuitée.
source

MCP Tool-call Latency

DIFFERENTIATOR· sample data · production shape
p50 / p95 / p99 latency per MCP endpoint (search · rates · book · static · vouchers · loyalty · flights). Shape comes from the open-source liteapi-travel/mcp-server; timings here are simulated for the demo.
MCP tool-call successi
98.10%
p50 latency · /searchi
108ms
p95 latency · /hotels/ratesi
412ms
Tool-call sequence lengthi
3.4
MCP endpoint latency · last 30 days (p95)
four busiest tools · timing middleware in a fork of liteapi-travel/mcp-server
0ms117ms234ms351ms468msD1D5D9D13D17D21D25D29
/search/hotels/rates/bookings/flights
Success rate per MCP tool
last 7 days, sorted worst-first
/search
99.73%
/loyalty
98.89%
/bookings
98.17%
/flights
97.96%
/static
97.73%
/hotels/rates
97.63%
/supplyCustomization
97.49%
/vouchers
97.22%
p95 latency per MCP tool
last 7 days, sorted slowest-first
/hotels/rates
412ms
/flights
318ms
/bookings
224ms
/search
168ms
/static
142ms
/loyalty
96ms
/vouchers
74ms
/supplyCustomization
58ms
Why this is a differentiator OPEN-SOURCE WEDGE
No competitor publishes per-tool MCP latency, because no competitor open-sourced an MCP server. Forking liteapi-travel/mcp-server and adding a 30-line timing middleware lets you expose these dashboards publicly at, say, mcp-status.liteapi.travel — instant developer-trust signal + recruiting flywheel.
Sample data. Latency and success-rate numbers are deterministic synthetic data. Production version: fork the MCP server, add timing middleware, land timings into BigQuery via Vercel Edge logs (~2 days of work).