{ "suite": "executor-cmp-latency-chart", "mode": "equal", "canonical": "https://clawql.com/benchmarks/executor-comparison/latency.html", "measuredAt": "2026-10-11T00:31:37.146Z", "host": { "node": "v22.14.0", "platform": "linux", "arch": "x64" }, "config": { "iters": 100, "warmup": 10, "mock_delay_ms": 0, "pet_count": 2, "body_bytes": 98, "equal_payload": { "pets": [ { "id": 1, "name": "Ada", "status": "available" }, { "id": 2, "name": "Grace", "status": "available" } ] }, "fields": [ "pets" ], "where": null }, "pathFlags": { "CLAWQL_TIER": "gateway", "CLAWQL_CAPABILITY_LIFECYCLE": "0", "CLAWQL_WORM_ENABLED": "unset", "panguard_sidecar": false, "durable_worm": false, "ephemeral_ring_audit": true, "hook_registry_pre_call": true, "where_and_fields": "fields_only", "mode": "equal" }, "workloadTilt": { "mode": "equal", "clawql": "EQUAL — single MCP execute → tiny same-host mock (2 pets), fields projection only. No search bundle, no where, no WORM, no audit in the timed execute sample.", "executor": "EQUAL — single MCP execute returning the same pets JSON in-process (Executor sandbox cannot fetch; logical result matched).", "intent": "Apples-to-apples: same result shape, one MCP execute each. Headline uses ClawQL gateway overhead (execute − direct) vs Executor execute." }, "tools": [ { "id": "clawql_execute", "source": "live", "label": "ClawQL execute (e2e)", "workload": "equal", "subtitle": "MCP → tiny same-host mock + fields", "p50_ms": 1.753, "p95_ms": 2.691, "p99_ms": 4.159, "p999_ms": null, "min_ms": 1.28, "max_ms": 158.237, "n": 100 }, { "id": "clawql_gateway_overhead", "source": "derived", "label": "ClawQL gateway overhead", "workload": "equalized", "subtitle": "execute − direct fetch (apples-to-apples)", "p50_ms": 1.341, "p95_ms": 2.147, "p99_ms": 3.399, "p999_ms": null, "headline": true }, { "id": "executor_execute", "source": "live", "label": "Executor execute", "workload": "equal", "subtitle": "live · same pets JSON in-process", "p50_ms": 3.032, "p95_ms": 5.512, "p99_ms": 5.88, "p999_ms": null, "min_ms": 2.281, "max_ms": 7.431, "n": 100, "endpoint": "stdio:/tmp/executor-try/node_modules/.bin/executor mcp", "citation": "Equal arm: MCP execute returns the same pets JSON ClawQL gets from the mock. In-process (Executor has no fetch); ClawQL pays same-host mock HTTP (~subtracted in overhead).", "headline": true }, { "id": "direct_http", "source": "live", "label": "Direct HTTP mock", "workload": "control", "subtitle": "bare fetch of the same tiny body", "p50_ms": 0.412, "p95_ms": 0.544, "p99_ms": 0.76, "p999_ms": null, "min_ms": 0.321, "max_ms": 1.899, "n": 100 }, { "id": "clawql_audit", "source": "live", "label": "ClawQL audit", "workload": "control", "subtitle": "local ring append (no upstream)", "p50_ms": 0.589, "p95_ms": 14.105, "p99_ms": 31.184, "p999_ms": null, "min_ms": 0.287, "max_ms": 190.424, "n": 100 } ], "series": { "clawql_execute": { "source": "live", "label": "ClawQL execute", "p50_ms": 1.753, "p95_ms": 2.691, "p99_ms": 4.159, "p999_ms": null, "min_ms": 1.28, "max_ms": 158.237, "n": 100, "citation": "MCP stdio → listPets equal tiny mock" }, "clawql_gateway_overhead": { "source": "derived", "label": "ClawQL gateway overhead", "p50_ms": 1.341, "p95_ms": 2.147, "p99_ms": 3.399, "citation": "execute − direct_http (same mock)" }, "clawql_audit": { "source": "live", "label": "ClawQL audit", "p50_ms": 0.589, "p95_ms": 14.105, "p99_ms": 31.184, "p999_ms": null, "min_ms": 0.287, "max_ms": 190.424, "n": 100 }, "direct_http": { "source": "live", "label": "Direct HTTP", "p50_ms": 0.412, "p95_ms": 0.544, "p99_ms": 0.76, "p999_ms": null, "min_ms": 0.321, "max_ms": 1.899, "n": 100 }, "executor": { "id": "executor_execute", "source": "live", "label": "Executor execute", "workload": "equal", "subtitle": "live · same pets JSON in-process", "p50_ms": 3.032, "p95_ms": 5.512, "p99_ms": 5.88, "p999_ms": null, "min_ms": 2.281, "max_ms": 7.431, "n": 100, "endpoint": "stdio:/tmp/executor-try/node_modules/.bin/executor mcp", "citation": "Equal arm: MCP execute returns the same pets JSON ClawQL gets from the mock. In-process (Executor has no fetch); ClawQL pays same-host mock HTTP (~subtracted in overhead)." } }, "headline": { "clawql_p50_ms": 1.341, "clawql_p95_ms": 2.147, "clawql_p99_ms": 3.399, "clawql_p999_ms": null, "clawql_arm": "gateway_overhead", "clawql_execute_e2e_p50_ms": 1.753, "executor_p50_ms": 3.032, "executor_p95_ms": 5.512, "executor_p99_ms": 5.88, "executor_p999_ms": null, "ratio_executor_over_clawql": 2.26, "ratio_clawql_over_executor": 0.44, "executor_source": "live", "clawql_faster_p50": true, "p50_beats_executor_band_low": true, "p99_beats_executor_band_high": true, "p999_beats_executor": false }, "honesty": { "flamegraphDemo120ms": "Synthetic fixture in clawql-observability demo-mcp-execute — NOT a measured ClawQL p50.", "tokenBenchmarks": "docs/benchmarks/executor-comparison/executor-cmp-*.json measure tokens, not wall clock.", "equalArms": "Both arms return the same tiny pets JSON via one MCP execute. ClawQL goes through same-host mock HTTP; that cost is subtracted for the equalized overhead metric.", "notHeavyTilt": "Previous heavy-tilt chart (800-row where/fields + search/WORM + audit) was retired for this page.", "referenceBand": "Do not use UsefulSoftwareCo/executor#1519 50–100ms as measured — wire EXECUTOR_BIN for live.", "p99Caveat": "p99 needs adequate n; default iters=100. Still sensitive to rare GC/scheduling spikes.", "executorReference": "Executor is LIVE equal-arm MCP execute returning the same pets JSON.", "notTokenFlamegraph": "Distinct from /mcp-ui/trace/compare/executor (token context). This page is wall-clock ms.", "p99VsExecutor": "Equalized ClawQL gateway overhead p50 1.341ms beats live Executor p50 3.032ms on the same result shape." } }