Every model, on a budget you control.
MeshPool is the gateway your company buys its AI through. The same models the labs sell, priced under their own list prices, with a budget each person and team sets for itself. Put in what you spend today and see what you keep.
$10 of free credit for every company that joins. One endpoint for every lab, one invoice for the company.
What I spend on AI today
What I want for that money
What you keep
$758
saved per month · $9,096 a year
Best models in this mix
31
models, every major lab, one endpoint
100%
of spend attributed to a person or service
2
views of spend: people and applications
$0
platform fee. You pay for tokens, nothing else
What you get
Spend decided by the people doing the spending.
Your own budget
Every person, team and service holds a budget it can see and set: a ceiling by day, week or month. When it runs out the calls pause instead of quietly adding to an invoice.
Your own models
Pick what you want from the whole catalogue and keep the rest out of your way. Every lab is reachable through the one endpoint, with no second contract to sign.
Your own way to spend
Say what the budget is for: cheapest that does the job, balanced, or best available. Give a single model its own sub-budget when it is the one that gets expensive.
04
Insights in production
Once traffic is real, spend is attributed to every key and model, with the change worth making named beside it.
See the insights →Budgets
Everyone sets their own ceiling.
Finance sets the company envelope once. Inside it, each person, team and service decides its own budget, its own models and how it wants to spend. Nobody waits on a ticket to try a model, and nobody finds out the number at the end of the month.
Insights in production
Where the spend goes, and what holds it in place.
Spend is attributed to the key that made the call, so what people run and what your applications run never arrive as one lump.
People
+8% vs August
$5,180
312 keys · 57% of spend
Applications
+31% vs August
$3,940
11 keys · 43% of spend
Spend
Every call attributed before it is billed
Cost lands against the key, the person and the model that made it, in the same window the call happened.
Controls
Ceilings, rate limits and model access
Set company-wide, per team, or on one account. A seat override beats the team, and the team beats the company list.
Guardrails
Checked before the call leaves the gateway
Budgets are enforced ahead of the lab, and secrets and personal data are redacted or refused on the way out.
Nothing switches on its own. Every number here is measured on your own traffic, and every control stays where an owner set it.
Pricing
Cheaper API keys, and no platform fee.
The keys you buy through the gateway are priced under the labs’ published list prices, and the platform itself costs nothing. You pay for tokens, and budgets, model picking and insights come with them.
Set your first budget in a minute.
Every company that joins starts with $10 of free credit. Sign in with the identity you already use at work, pick your models, set the number you are willing to spend. Rate-card pricing applies from the first token.
Already have a workspace? Sign in. Rate card version 2026-09-23.2.