Agent Runtime Router
Loading local engines

Binary workload placement

Route agent work without guessing.

Declare capability boundaries, price a scenario, and let the solver assign every task exactly once.

Rates are scenario inputs in micro US dollars per second.

The quoted $0.00001/s x86 rate is prefilled as 10 microUSD/s. It is source context, not verified current pricing.

Run a scenario to populate both runtime lanes with verified binary assignments.

Routing assumptions and unavailable local engines

Read the explanation

The displayed runtime assumptions enter two micro dollars per second for WASM and ten for x eighty six. One hundred active seconds would cost two hundred and one thousand micro dollars respectively before any startup charge. At point two pixels per micro dollar the bars span forty and two hundred. These are scenario inputs, not verified current pricing or measured runtime speed. The local snapshot contains only its HTML and lacks the referenced engine and app scripts, so Run Bundled Sample cannot populate assignments. We preserve that original inactive engine state and do not claim a solver result or actual runtime execution. A half-dollar budget equals five hundred thousand micro dollars. Dividing by the entered rates gives two hundred fifty thousand WASM seconds or fifty thousand x eighty six seconds if all that budget buys steady-state runtime and startup is ignored. At point zero zero one pixels per hypothetical second the bars span two hundred fifty and fifty. Actual workload eligibility, task counts, startup overhead, and required capabilities would change that accounting. A cheaper rate does not mean every task can run there. The source explicitly says capability constraints are evaluated before cost; its missing engines prevent checking that promise locally. The page declares that every task should be assigned exactly once within capability boundaries. For a hypothetical five-task workload the total assignment count must therefore be five. At forty pixels per assignment that spans two hundred, compared with one hundred sixty for only four assignments. Counts alone still cannot prove no task was duplicated while another was omitted; task identifiers and binary placement need independent inspection. The displayed capability fields use zero or one, but no parsed CSV, optimized JSON export, or budget-feasibility decision is demonstrated while the original local scripts are unavailable. This explains declared expectations without replacing the absent model or solver.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.