Roo Code API relay on Vana: $0.1389 per million input tokens

§ One address, three protocols, minimal setup

Vana API runs on a single address, https://api.vanaapi.com, which answers API requests. The dashboard lives separately at https://vanaapi.com, where you create keys, check your balance, and read the usage log. This separation keeps the API endpoint clean while giving you a full account view in the browser.

Three protocols share that one address. Claude models speak only from the official Claude Code or Claude Desktop, version 2.1.285 or later, over Anthropic Messages at /v1/messages; they do not work from any other client, SDK, cURL or script. Gemini models work only from Gemini CLI over the native protocol at /v1beta/models/..., and they are not OpenAI-compatible. The remaining models, including GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo, use the OpenAI format at /v1/chat/completions, which is the format Roo Code, Cursor, Cline, OpenCode and Cherry Studio already speak. GPT models also answer on /v1/responses, which is what Codex uses.

Because all three protocols sit on https://api.vanaapi.com, you do not maintain separate endpoints for different tools. You pick the client that matches the model family, point it at the one address, and Vana API routes the request correctly. That is the minimal setup: one address, three protocols.

§ A five-minute quick start with Roo Code

Roo Code pairs with the OpenAI-format models on Vana API, including GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo. To begin, open the dashboard at https://vanaapi.com and create an API key. Your sign-up includes $0.10 of credit, so you can send a test request before adding any funds. No monthly fee applies at this stage; you pay only for the tokens you actually use.

In Roo Code, set the base URL to https://api.vanaapi.com and paste the key you created. Choose a model id from the price list, for example gpt-6.1-sol, deepseek-v4-flash, grok-4.7, or qwen3.8-max. The client sends its requests to /v1/chat/completions, and Vana API forwards them to the chosen model over the same address. You do not need to configure a different endpoint for each model, and you do not need to install a proxy or a wrapper.

Roo Code is one of several clients that speak the OpenAI format. Cursor, Cline, OpenCode, Cherry Studio, any OpenAI SDK, and even cURL all work against the same /v1/chat/completions endpoint on https://api.vanaapi.com. That means the address you configure once in Roo Code also serves your other tools, so you are not locked into a single editor. The model id you pass in the request body determines which model answers; Vana API does not require a different base URL per model.

The whole process fits inside five minutes. Create the key in the dashboard, set the base URL in Roo Code, pick a model id, and send your first message. If you later want to try Codex, the same address answers on /v1/responses for GPT models, so the setup carries over without a new endpoint.

§ Pricing against official list prices

Vana API charges per real token from a prepaid balance, with no monthly fee unless you choose a plan. Every call is listed in the usage log in the dashboard, so you can see exactly what each request cost. For GPT-6.1 Sol, model id gpt-6.1-sol, Vana API lists $0.1389 per million input tokens and $0.8333 per million output tokens, while the official list price is $2 / $10. For GPT-5.6 Sol, model id gpt-5.6-sol, the Vana API price is $0.3472 / $1.81 against an official $4 / $20. For DeepSeek V4 Flash, model id deepseek-v4-flash, the price is $0.2778 / $1.04 per million tokens, and for Grok 4.7, model id grok-4.7, it is $0.4861 / $1.53.

For example, MiniMax M3, model id MiniMax-M3, is listed at $0.1389 / $0.5556 per million tokens, and MiMo V2.6 Pro, model id MiMo-V2.6-Pro, at $0.2083 / $0.3472. Qwen3.7 Plus, model id qwen3.7-plus, costs $0.1278 / $0.4167. These are real numbers from the current price list, and they apply per real token, not per request. If you switch models mid-project, the same address and key continue to work; you only change the model id in your client.

Say you run a coding session that consumes one million input tokens and half a million output tokens on gpt-6.1-sol every day; at Vana API prices that example costs $0.1389 for input plus about $0.4167 for output. Over a month, the same example would total roughly $4.17 for input and $12.50 for output. Compare that with keeping an official account or a fixed subscription, where the same volume might exhaust a monthly allowance or cost more per token. With Vana API you stop paying when you stop calling; there is no subscription to cancel.

§ The dashboard, keys, and the usage log

All account management happens at https://vanaapi.com, not at the API address. There you create keys, top up your prepaid balance, and review the usage log, which lists every call with its token count and cost. This log is the single source of truth for spending; you do not need a separate invoice or a third-party tracker.

The dashboard at https://vanaapi.com is also where you manage keys. You can create multiple keys for different projects, each with its own usage tracking in the log, and revoke them when a project ends. Because the API address itself only answers requests, it stays lean; all the account logic lives in the dashboard.

Refunds are available for unused balance within 3 days of a top-up, minus what was used and the payment processor's fee. If you invite others to Vana API, you earn 14% of what they pay, and 5% one level further; those earnings can be withdrawn from $13.89. The $0.10 sign-up credit lets you verify that your Roo Code configuration works before you commit larger amounts.

§ When a monthly plan fits

For steady, GPT-only workloads, Vana API offers a monthly plan of $103.61 for 30 days, covering up to $4,000 in usage over those 30 days, with at most $1,200 in any single week. This plan applies to GPT models only, and it is optional; the prepaid balance remains the default for everyone else. If your monthly usage stays below the $4,000 ceiling, the plan fixes your cost for the month at $103.61. If your usage is lighter or irregular, the pay-as-you-go balance keeps you paying only for real tokens, with no monthly fee and no commitment.