Design-partner program
Run your agents on your own keys, with us in the room.
We are taking five teams that run AI agents in production. You get hands-on setup and first access to what we build next. We get real workloads and a published account of what happened, with numbers you sign off on.
Who it fits
- Agents in production: scheduled jobs, sub-agent fleets, coding agents or customer-facing assistants.
- Calls to more than one model provider, or a plan to add one.
- Model spend somewhere around $1,500 to $50,000 a month.
- You want spend caps and failover on your own provider accounts, without a prepaid balance.
What you get
Hands-on setup
A working session with the engineer who built ROUTEXOR to set up your task profiles, fallback chains and spend caps, then a direct line while you run.
Your issues go first
Bugs and missing pieces you hit get worked before anything else on the roadmap.
First to test per-run budgets
Cumulative budgets per agent run are built and switched off today. We will turn them on with design partners first, after checking them against your workflow together.
Adapters for what you run
Framework integrations get built in the order our partners need them.
What we ask
Real traffic for 30 days
Route at least one production agent through ROUTEXOR for 30 days or more.
A short check-in every two weeks
30 minutes on what worked, what broke and what is missing.
A one-page case study
Published with your numbers, in the format below. You approve it first, and it can be anonymous.
Terms
Design partners use the same public plans as everyone else: Free at $0 with 1,000 requests a day, Starter at 2.2% of metered usage, Pro at a flat $99 a month, or Enterprise by agreement. Most partners can start on Free and move up only if their volume needs it.
Your provider keys stay yours and your providers keep billing you directly. Leaving means deleting your keys from the vault and pointing your agents back at your providers. There is nothing to migrate and no balance to recover.
Before you decide, read the security and key-management whitepaper and check the live status page.
Case studies
ATERNA's agent fleet on ROUTEXOR
ATERNA builds ROUTEXOR, so this is not a customer. It is the one production workload we can show today: 34,597 requests and 835.7M tokens routed as of October 12, 2026.
One-page case-study template
A 30-day baseline from your invoices, the same 30 days after the switch, and a source for every number. You approve it before it is published.
Apply
Email team@aterna.ai with "ROUTEXOR design partner" in the subject. Tell us what your agents do, which providers and models they call, your rough monthly model spend and what you most want to control.