Roo Code pay-as-you-go pricing at Vana API starts $0.10

§ Why does Vana API run on a single address for Roo Code

Vana API operates from one address, https://api.vanaapi.com, which handles three different protocols behind the same URL. For developers using Roo Code, the relevant path is the OpenAI-compatible format at /v1/chat/completions, so the client talks to Vana API exactly as it would to any other OpenAI-format endpoint. You do not configure multiple hosts or switch addresses when you change models. The management side lives separately at https://vanaapi.com, where you create API keys, watch your prepaid balance, and read the usage log. That API address only answers requests and never serves a dashboard. This split keeps the setup minimal: one place to make calls, one place to administer them.

The practical benefit for an indie developer is a calm, predictable workflow. When Roo Code sends a request to https://api.vanaapi.com/v1/chat/completions, Vana API routes it to the model you selected, such as Qwen3.7 Plus or GPT-6.1 Sol, and returns the response in the exact format Roo Code expects. You do not maintain separate configurations for different tools or remember which endpoint belongs to which client. The single address also means your billing stays unified, because every call lands in the same usage log on the dashboard regardless of which OpenAI-compatible model processed it. This removes the mental overhead of tracking multiple services.

§ Which models can Roo Code use through Vana API and what do they cost

Roo Code works with the OpenAI-format models that Vana API serves: GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax, and MiMo. Each model has a specific model id you enter in Roo Code, and the pricing is per real token with no monthly fee. For example, Qwen3.7 Plus (model id qwen3.7-plus) costs $0.1278 per million input tokens and $0.4167 per million output tokens. GPT-6.1 Sol (model id gpt-6.1-sol) costs $0.1389 per million input and $0.8333 per million output, compared with an official list price of $2 and $10. DeepSeek V4 Flash (deepseek-v4-flash) is $0.2778 / $1.04, MiniMax M3 (MiniMax-M3) is $0.1389 / $0.5556, and Grok 4.7 (grok-4.7) is $0.4861 / $1.53.

Say you run Qwen3.7 Plus every day for local coding tasks. Because Vana API charges only for the tokens actually processed, your cost for a session depends directly on how much text the model reads and writes. There is no subscription to cancel and no minimum spend beyond the prepaid balance you choose. If your project shifts to a heavier model like Kimi K3 at $1.04 per million input and $5.21 per million output, you simply change the model id in Roo Code; the address https://api.vanaapi.com stays the same. You can also use GLM-5.3 at $0.5556 / $1.67, Qwen3.8 Max at $0.6944 / $2.08, or MiMo V2.6 Pro at $0.2083 / $0.3472. All of them answer on the same /v1/chat/completions endpoint, so switching in Roo Code is just a matter of changing the model id string.

§ What does the setup process look like from sign-up to first call

Getting started with Vana API for Roo Code takes a few minutes and requires no additional software beyond Roo Code itself. First, you visit https://vanaapi.com and create an account. New accounts receive $0.10 of credit, which is enough to run several test calls with a smaller model like Qwen3.7 Plus or MiniMax M3. In the dashboard on that same site, you generate an API key and note your current balance. The key is the only credential you need to authenticate requests.

Next, you open Roo Code and set the base URL to https://api.vanaapi.com. You paste the API key into the authentication field and choose a model id from the OpenAI-compatible list, such as gpt-6.1-sol, deepseek-v4-flash, or glm-5.3. No additional proxies, SDK wrappers, or protocol converters are required for these models. When you send your first message, Roo Code posts to /v1/chat/completions at that address, and Vana API responds in the standard OpenAI format. The call appears immediately in the usage log on https://vanaapi.com, so you can verify that the billing matches the tokens shown in Roo Code's own interface.

§ How does pay-as-you-go billing work at Vana API

Vana API uses a prepaid balance. You top up the balance on https://vanaapi.com, and each API call deducts the exact cost of the tokens consumed at the published rates. There is no monthly fee and no commitment; you stop spending when you stop calling. Every request is recorded in the usage log in the dashboard, so you can audit each call against your balance in real time and see exactly which model id generated which charge.

For heavier workloads that stay within the GPT family, there is a monthly plan option. You can pay $103.61 for 30 days of access that covers up to $4,000 of GPT model usage, with a weekly cap of $1,200. This plan applies to GPT models only and runs through the same single address, https://api.vanaapi.com. If you prefer to stay fully variable, the standard prepaid model remains available with no cap other than your balance. The per-token model means you are never charged for idle time. Unused balance can be refunded within 3 days of the top-up, minus what was already used and the payment processor's fee, giving you a short window to adjust if you overfunded the account.

§ How can you track spending and earn through the referral program

The dashboard at https://vanaapi.com shows every call in the usage log, including the model, the token counts, and the amount deducted. This transparency lets you compare costs across models; for instance, running MiniMax M3 at $0.1389 per million input tokens versus Kimi K3 at $1.04 per million input tokens shows a clear difference in daily spend for the same task. Because the log updates with each request, you always know your remaining balance before starting a long session.

Vana API also runs a referral program. When you invite others, you earn 14% of what they pay toward their prepaid balance, and 5% from users one level further down. Earnings can be withdrawn starting from $13.89. Because the service bills per real token and lists every transaction in the usage log, you can see exactly how referral income accumulates alongside your own usage, and the same dashboard manages both your keys and your earnings.