Welcome to ModelRig

See every model call, grade what's working, and improve cost and quality on the ones that matter.

Get started

Finish setting up your rig

Mint an API key, point your coding agent (or your app) at ModelRig, and make one call. It takes about two minutes, and every screen below fills in from there.

Open setup & onboarding

See

Start here — the truth of what your calls actually did.

Calls

Every call, with its cost, latency, and model.

Runs

Grouped calls with their steps and grades.

Route readiness

How much evidence each route has to optimize safely.

Action queue

What the Rig Analyst found across your traffic.

Grade

Judge quality — a 0–1 score plus a comment, from your code, a human, or an AI.

Graders

Teach an AI grader with a guide, and check its calibration.

Review

Read a bake-off's outputs side by side and grade them.

Grades

Every grade, and which routes still need grading.

Improve

Bake off a cheaper-or-better model and swap — with the evidence.

Bake-offs

Head-to-head model comparisons on your tasks.

Proposals

Swap and routing changes, with the evidence.

Leaderboard

Declared vs. probed, priced, ranked.

Spend

Cost, and per-customer invoices.

Costs

Spend by model, route, and any tag you attach.

Tenants

A per-customer statement: attempted, verified, billable.