Service 05 — one endpoint, every model
One endpoint. Every model. Your keys.
Hosted gateways take up to five percent of everything you spend on tokens — a tax that scales with your success. Ours is a flat monthly fee on infrastructure we manage: your keys, your data path, provider failover, spend budgets, full request logs.
Managed by StanServe · billed in USD via Stan Billing
Plans & pricing
Sized honestly, priced in USD
Every plan is fully managed — monitoring, patching and verified backups are the baseline, not an add-on.
Route One
For one product finding its feet
$49 USD/mo
per month · 0% token markup
- Managed gateway (shared HA cluster)
- Bring your own provider keys
- Failover across providers
- Spend budgets + alerts
- 30-day request logs
Route Pro
For teams shipping on AI
$149 USD/mo
per month · 0% token markup
- Dedicated gateway instance
- Per-team keys and budgets
- Prompt/response caching
- Latency and cost analytics
- 90-day request logs
Route HA
For AI in the critical path
$449 USD/mo
per month · 0% token markup
- Active-active HA pair
- Private endpoints / VPN peering
- Compliance-grade audit logging
- Custom retention policies
- Priority response window
The math
Percentage vs. flat
On $10k/month of tokens, a 5% platform fee is $6,000 a year. Our biggest router plan costs less than that — and the fee never grows with your success.
Hosted gateway, % fee
- Fee scales with every token you spend
- Your traffic transits their infrastructure
- Their keys or theirs-wrapped-yours
- Rate limits you share with strangers
StanServe router, flat fee
- Flat monthly fee, whatever you spend
- A gateway instance that is yours
- Your provider keys, held in a vault
- Your limits, your budgets, your logs
What the router does
The connectivity layer for AI
One API, every provider
Anthropic, OpenAI, Google, open models — one endpoint and one bill of logs, whatever sits behind it.
Failover that just happens
A provider outage becomes a retry against the next model on your list, not a pager alert.
Budgets with teeth
Per-team and per-app spend caps that actually stop requests, not dashboards you check after the damage.
Every request logged
Prompts, latencies, costs — queryable, exportable, yours. Compliance stops being a spreadsheet.
Model routing
Stop paying a percentage to reach your own models.
Point one staging app at a router for a month and read the logs. The case makes itself.