Hard ceilings
A tenant that hits its budget gets a clean 429, not a surprise invoice.
01Cost control for inference workloads
Per-tenant budgets with hard ceilings, soft landings and live attribution by feature.
Erekm gives your team one place to cost control for inference workloads — with limits that are visible before they are hit.
No card required. 38% spend cut in one quarter.
Overview
Runs
Policies
Traces
Settings
Costlive
38%spend cut in one quarter
0invoice surprises since launch
Hard ceilingsok
Attribution by featureok
Forecast that warnsok
Teams shipping with Erekm
NorthwindKestrelRuangPolarTokoLumenBakauVela
38%spend cut in one quarter
0invoice surprises since launch
12providers in one view
What you get
Three things Erekm does properly.
Attribution by feature
Every token is tagged with the feature that spent it, down to the endpoint.
02Forecast that warns
Projected spend crosses a threshold and the right person hears that day.
03Inside the product
One screen, the whole picture.
Overview
Runs
Policies
Traces
Settings
Costlive
38%spend cut in one quarter
0invoice surprises since launch
Hard ceilingsok
Attribution by featureok
Forecast that warnsok
“We found two features burning 60% of the budget. Nobody had noticed, because nobody owned the number.”
Kuba M. — SRE, Warsaw
Pricing
Start free. Pay when it earns it.
Questions
The things people ask first.
Do you proxy traffic?+
Only if you want us to. Metering works from logs too.
Which providers?+
OpenAI, Anthropic, Google, Bedrock, Together, Fireworks, vLLM, and your gateway.
Budgets per customer?+
That is the point. Tenant, feature, model and environment are all dimensions.
Erekm is open for design partners.
Small team, direct access to the people who build it. Bring a real workload.
Request access