GatewayBench
GatewayBench: Unified Capability and Security Evaluation Platform for MCP Gateways
Evaluates MCP gateway architectures on capability, efficiency, and security using 524 tasks, 11k+ tools, and verifiable judges.
How to run
$ npm i -g lettokenburn-cli$ ltb bench info gatewaybench
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
3 sourced entriesEvery number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsAbout
GatewayBench comprises 524 tasks across six domains with a pool of 11,118 real-world tools from 251 MCP servers. It injects task-irrelevant noisy tools to simulate large-scale deployments.
Scoring uses verifiable judges to inspect environment states quantitatively. Key findings reveal hierarchical gateways offer better scalability and security against indirect attacks compared to flattened ones.