The honest details
behind every tier.
How per-application billing works, what fair-use API rate limits actually mean, and how to bring your own AI. No usage metering, no surprise line items.
Every plan, side by side
One bill: your deployed applications. Bring your own AI on every tier — Free through Enterprise.
| Free$0/mo | Team$49/moMost popular | Business$149/mo | EnterpriseCustom | |
|---|---|---|---|---|
| Applications included | 1 | 5 | 20 | Unlimited |
| Each additional app | — | +$8 / app / mo | +$6 / app / mo | Included |
| API rate limits | Fair use | Higher | Unlimited (fair use) | Unlimited, dedicated |
| MCP endpoints | 1 | Per app | Per app | Per app + custom |
| Glass Box tracing | Basic | Standard | Advanced | Full audit suite |
| SSO, roles & audit trails | — | |||
| Team members | 1 | Up to 3 | Up to 15 | Unlimited |
| Bring your own AI | + BYOM | |||
| Self-hosted on your K8s | — | — | — | |
| Support | Community | Priority + SLA | Dedicated + SLA | |
| Start free → | Get Team → | Get Business → | Contact sales → |
You're billed per application
A deployed app is its own MCP surface, routes, data, and Glass Box audit trail. Each plan includes a set number; add more at a flat monthly rate per extra app. That's it — no usage metering, ever.
| Plan | Applications included | Each additional app |
|---|---|---|
| Free | 1 | Not available — upgrade to add |
| Team | 5 | $8 / app / mo |
| Business | 20 | $6 / app / mo |
| Enterprise | Unlimited | Included |
What "fair-use API rate limits" means
OBTO meters application hosting — not your model calls. Rate limits keep the shared platform healthy; they are not a billing meter.
Limits, not meters
There's no hard per-call counter on your bill. Fair use means we only step in to throttle abnormal or abusive automated load — the kind that would degrade the platform for everyone.
Room to grow
- Free: fair-use limits, sized for one prototype app
- Team: higher limits across your apps
- Business: unlimited calls under fair use
- Enterprise: unlimited on dedicated infrastructure
We reach out first
If you consistently push the limits, we contact you to move to the right plan — we never let a runaway loop turn into a surprise invoice, because there's no per-call charge to spike.
Two ways to connect your model
Either path keeps your LLM spend with your own provider. OBTO hosts the app and never routes your inference.
Point your MCP client at OBTO
- Sign in to OBTO to reach your MCP endpoint
https://app.obto.co/ms/mcp - Add it to Claude Desktop, Cursor, ChatGPT, or any MCP-compatible client
- Ask your AI to build & deploy — it runs on its own model, OBTO hosts the app
Your client talks to its own LLM. OBTO only sees tool calls — never your inference.
Run it with your own key
- Choose your provider — OpenRouter, Anthropic, or OpenAI
- Add your API key to your OBTO workspace settings
- OBTO calls that endpoint with your key; the model bill lands on your provider account
Ideal for headless agents and scheduled jobs where there's no interactive client in the loop.
Provider spend is always yours. Whether you connect a client or bring a key for OpenRouter, Anthropic, or OpenAI, your model usage is billed by that provider directly. OBTO's invoice only ever contains your deployed applications.
See exactly what your agents did
Every plan traces each request end to end. Higher tiers add retention, depth, and audit reach.
Basic tracing
Per-request traces for your single app: tool calls, inputs, outputs, and errors, so you can debug what happened.
Standard tracing
Richer traces with longer retention and team-wide visibility across your apps, plus SSO, roles, and audit trails.
Advanced tracing
Advanced traces, extended retention, and deeper audit for production workloads running at scale.
Enterprise — full audit suite. Everything in Advanced, plus the complete Glass Box audit suite, custom retention, and self-hosted deployment on your own Kubernetes so traces never leave your infrastructure.
Help that scales with the plan
| Plan | Support | Team & access |
|---|---|---|
| Free | Community support | 1 member |
| Team | Email support | Up to 3 members · SSO, roles & audit trails |
| Business | Priority support + SLA | Up to 15 members · SSO, roles & audit trails |
| Enterprise | Dedicated support + SLA | Unlimited members · custom MCP integrations |
The questions we get most
What counts as an app?
A deployed OBTO app with its own MCP surface, routes, data, and Glass Box audit trail. Many pages and routes can live inside one app — the bill is per deployed app, not per page or feature.
What happens over fair-use?
Nothing surprising. There's no per-call charge to spike — we only throttle abnormal or abusive automated load, and we reach out to move you to the right plan if you consistently push the limits.
Can I self-host?
Yes, on Enterprise — the full OBTO runtime on your own Kubernetes cluster, with your own models and custom MCP integrations, backed by dedicated support and an SLA.
Do you ever charge for model usage?
No — never. Bring your own AI, always. Connect a client or bring an OpenRouter, Anthropic, or OpenAI key; your model spend stays with your provider. OBTO only bills per deployed app.
Bring your own AI. Ship on OBTO.
Start free with one app, or jump into Team and scale. You keep your LLM spend; we host the app.