Roo Code setup on Vana API: $0.10 credit gift

§ One address, three protocols, minimal setup

Vana API is built around a single technical address. Every API request goes to https://api.vanaapi.com. That one endpoint speaks three protocols, so you do not need separate hosts for different tools. The dashboard, where you create keys, check your prepaid balance and read the usage log, lives at https://vanaapi.com.

Roo Code is an OpenAI-format client. It sends its requests to /v1/chat/completions, which is one of the three protocols that Vana API exposes on that single address. Because Roo Code already speaks this format natively, connecting it requires no protocol translation, no wrapper and no extra middleware.

The setup is minimal: one base URL, one key, one model id. You can keep this configuration and switch between compatible models by changing only the model id, without touching the address or the protocol settings.

§ Create your key and use the sign-up credit

Start at https://vanaapi.com and register. On sign-up you receive $0.10 of credit automatically. This is real balance, not a trial flag; it can be spent on any available model at the listed per-token prices. In the dashboard you generate your API key and add a prepaid balance if you want to go beyond the gift.

Billing is prepaid and charged per real token, with no monthly fee. You are not billed for idle time or for having a key; you pay only when a request runs and tokens are consumed. Every call is recorded in the usage log, so the balance and the log always match the actual usage.

Keep your key in the dashboard. The balance acts as a hard ceiling: once it is spent, requests stop, which makes spending easy to control without any subscription commitment.

§ Configure Roo Code once and follow the checklist

Open Roo Code and set the base URL to https://api.vanaapi.com. Paste the API key you created in the Vana API dashboard. Then pick a model id from the OpenAI-format list, such as gpt-5.6-sol, gpt-6-astra, deepseek-v4-pro, grok-4.7, glm-5.3, kimi-k3, qwen3.7-plus, qwen3.8-max, MiniMax-M3 or MiMo-V2.6-Pro. Roo Code will send its chat completion requests to that single address, and Vana API will route them to the chosen model.

There is nothing else to install. Because Roo Code uses /v1/chat/completions, it sits in the same protocol group as Cursor, Cline, OpenCode, Cherry Studio, any OpenAI SDK and plain cURL. You can switch models later by editing only the model id in Roo Code; the address and the key stay the same.

If you work with GPT models in Codex as well, note that Vana API also answers on /v1/responses for those, but for Roo Code the chat completions endpoint is the correct one to use. Follow this checklist once and the connection is ready.

§ Exact per-token prices for the models Roo Code uses

Vana API publishes prices per million tokens, split into input and output. For GPT-5.6 Sol (model id gpt-5.6-sol) the price is $0.3472 per million input tokens and $1.81 per million output tokens, against an official list price of $4 and $20. GPT-6.1 Sol (gpt-6.1-sol) is $0.1389 input and $0.8333 output. DeepSeek V4 Flash (deepseek-v4-flash) is $0.2778 input and $1.04 output, and DeepSeek V4 Pro (deepseek-v4-pro) is $1.11 input and $3.33 output. Qwen3.7 Plus (qwen3.7-plus) is $0.1278 input and $0.4167 output.

These are the exact numbers that appear in the usage log. Say you run this every day: a workflow that consumes one million input and one million output tokens on deepseek-v4-flash would cost $1.3178 for that day, calculated directly from the $0.2778 and $1.04 figures. Because billing is per real token, there is no rounding up and no minimum charge per request.

You can compare any model in the list before you commit. Grok 4.7 (grok-4.7) is $0.4861 input and $1.53 output, while MiniMax M3 (MiniMax-M3) is $0.1389 input and $0.5556 output. The choice is visible in the dashboard and in the log, so you always know which model produced which charge.

§ Watch the usage log and the refund window

Every API call is listed in the usage log on https://vanaapi.com. The log shows the timestamp, the model id, the token counts and the amount charged. Since Vana API bills per real token from a prepaid balance, you can reconcile every line against the price list and against your remaining balance without any separate invoice.

If you top up and then decide you do not need the full amount, unused balance can be refunded within 3 days of the top-up. The refund subtracts what was already used and the payment processor's fee, and the rest returns to you. This window is short, so check the log soon after a top-up if you want to adjust.

The combination of a prepaid balance, a per-token charge and a detailed log means the cost stays visible at all times. There is no monthly fee and no surprise overage; the balance simply decreases as Roo Code makes requests.

§ Optional extras: referrals and a monthly GPT plan

Vana API has a referral program. You earn 14% of what the people you invite pay, and 5% one level further down. Earnings can be withdrawn once they reach $13.89. The referral balance sits in the same dashboard as your main balance and usage log.

For predictable, heavier GPT-only workloads there is a monthly plan. It costs $103.61 for 30 days and covers up to $4,000 of GPT usage in that period, with a weekly cap of $1,200. The plan applies to GPT models only and is separate from the prepaid pay-as-you-go balance.

Both options use the same single address, https://api.vanaapi.com, and the same key you configured in Roo Code. The checklist you followed once is the whole setup; everything else is optional and visible in the dashboard.