Quick Start Guide

Paste a few real prompts and get a dollar figure in 5 minutes.

The fastest path from sign-up to a dollar figure you can defend. It takes about five minutes and does not require connecting anything.

1. Sign Up

Create an account and verify your email. An optional intake asks which catalog model serves production and roughly how many requests you make each month; those answers prefill later screens. Role, company, and additional context are optional and do not change the product flow. Every new organization gets one sponsored comparison on its own prompts.

2. Try a comparison

From the dashboard or the Monitors list, choose Try a comparison. Paste one to 20 prompts your product actually sends, tell us which model serves them today, and pick the challengers you are curious about.

Each example is judged by a five-model panel drawn from a frozen pool of eight, the same way a Monitor cycle judges production traffic. The sample is still small, so the result is labelled directional and tells you how many more examples would settle it — treat it as a reason to look closer rather than as a decision.

3. Read the Dollar Result

You get an estimated monthly spend, a projected saving, and the reasoning behind both. Costs are shown as a range at provider list price, stamped with the date those prices were read — never a single point estimate.

4. Turn It Into a Monitor

A one-off look on pasted examples cannot tell you whether a switch holds next week. The result page offers two ways forward. Add more examples reopens the comparison with your examples loaded, so you can paste more and rerun on the same Monitor. Connect production traffic asks where your traces live and opens the Monitor on that source: point an OpenTelemetry exporter at PeerLM, connect an observability tool you already run, or upload an export. Your examples, production model, and challengers carry forward either way.

5. Run a Comparison

With traffic flowing, a Run samples your prompts, replays each candidate against them, and seats a rotating panel of judges drawn from a pool frozen on the Monitor. PeerLM picks the judges — a model from the vendor under test never judges it. See Monitors overview.

6. Read the Verdict

Runs roll up into a living switch, route, or hold verdict on the Monitor, with the receipts on each Run. Every judged item is inspectable: the prompt, both outputs, the panel seated for it, and each judge's vote.