Claude Desktop and Vana API: one address relay service

§ One address, three protocols

Vana API serves every model from a single address, https://api.vanaapi.com. That one address accepts three protocols, so there are no separate URLs to remember for different tools. The setup stays minimal: you create an API key in the dashboard on the main site, https://vanaapi.com, and then you point each client at the same base address with that key.

Claude Desktop and Claude Code, version 2.1.285 or later, speak the Anthropic Messages protocol themselves and call /v1/messages on that address. They do not work from any other client. Gemini CLI speaks the Gemini native protocol and calls /v1beta/models/... on the same address; those models are not OpenAI-compatible. For GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo you use the OpenAI format at /v1/chat/completions, from Cursor, Cline, Roo Code, OpenCode, Cherry Studio, any OpenAI SDK or cURL. GPT models also answer on /v1/responses, which is what Codex uses.

The dashboard at https://vanaapi.com holds your keys, your prepaid balance and the full usage log. The API address itself only answers API requests and does not show the dashboard. When you sign up, Vana API adds $0.10 of credit to the new balance, so you can make a first call before any top-up. You manage keys and read the log on the main site, while every client sends its requests to https://api.vanaapi.com.

§ Claude Desktop on the one address

To run Claude models you use the official Claude Desktop or Claude Code, version 2.1.285 or later. You set the base URL to https://api.vanaapi.com and the client sends its requests over /v1/messages. No other client can reach the Claude models, so the path stays short: install the official client, set the address, enter the key from the dashboard.

Say you run this every day as an example. You pick Claude Opus 5.5, model id claude-opus-5-5, priced at $1.46 per million input tokens and $6.94 per million output tokens. If a working day uses 3 million input tokens and 1 million output tokens, the sum is 3 × $1.46 + 1 × $6.94 = $4.38 + $6.94 = $11.32. Vana API deducts $11.32 from the prepaid balance, and that call is listed in the usage log in the dashboard.

Say you run this every day as an example with a lighter model. Claude Haiku 4.5, model id claude-haiku-4-5-20251001, costs $0.3472 per million input tokens and $1.60 per million output tokens. A day with 20 million input tokens and 5 million output tokens costs 20 × $0.3472 + 5 × $1.60 = $6.944 + $8.00 = $14.944, charged per real token with no monthly fee. You can also compare Claude Sonnet 5.5 at $1.18 per million input tokens and $5.76 per million output tokens, or Claude Fable 5.1 at $3.33 and $16.74, all on the same address.

§ Your other tools sharing the same address

The same https://api.vanaapi.com address also serves your OpenAI-format tools. Cursor, Cline, Roo Code, OpenCode, Cherry Studio, any OpenAI SDK or cURL all point to that address and call /v1/chat/completions for GPT, Grok, DeepSeek, GLM, Kimi, Qwen, MiniMax and MiMo models. GPT models also respond on /v1/responses for Codex, so Codex works on the same base URL.

Say you run this every day as an example. In Cursor you call MiniMax M3, model id MiniMax-M3, at $0.1389 per million input tokens and $0.5556 per million output tokens. A session with 8 million input tokens and 3 million output tokens costs 8 × $0.1389 + 3 × $0.5556 = $1.1112 + $1.6668 = $2.778 from the same prepaid balance. Another example: GPT-6.1 Sol, model id gpt-6.1-sol, at $0.1389 per million input tokens and $0.8333 per million output tokens, for 10 million input and 2 million output tokens costs 10 × $0.1389 + 2 × $0.8333 = $1.389 + $1.6666 = $3.0556.

Gemini CLI connects to the same Vana API address over /v1beta/models/... for Gemini models. Say you run this every day as an example. Gemini 2.5 Flash, model id gemini-2.5-flash, is priced at $0.2083 per million input tokens and $1.60 per million output tokens. A day with 15 million input tokens and 4 million output tokens costs 15 × $0.2083 + 4 × $1.60 = $3.1245 + $6.40 = $9.5245, logged in the dashboard like every other call. You can also use Gemini 3.1 Pro at $1.39 and $7.92, Gemini 3.8 Flash at $1.04 and $4.65, or Gemini 4 Argon at $1.46 and $6.88, all through Gemini CLI on the one address.

§ Billing, gift credit and refunds

Vana API uses a prepaid balance and charges per real token. There is no monthly fee. Every call appears in the usage log in the dashboard, so you can see exactly what each request cost and which model it used. Keys are created in the dashboard at https://vanaapi.com. New accounts start with $0.10 of credit. If you top up and do not use the balance, you can request a refund of the unused amount within 3 days of the top-up, minus what was used and the payment processor's fee.

The referral program pays 14% of what the people you invite pay, and 5% one level further. Earnings can be withdrawn from $13.89. All of this sits on the same dashboard that shows the balance and the log, next to the single API address.

§ Monthly plans for steady use

If you prefer a fixed monthly amount instead of pure pay-as-you-go, Vana API lists three plans. The Claude plan up to $1,000 / 30 days costs $64.72 for 30 days and allows at most $320 a week, for Claude models only. The Claude plan up to $2,400 / 30 days costs $153.61 for 30 days and allows at most $750 a week, also for Claude models only. The GPT plan up to $4,000 / 30 days costs $103.61 for 30 days and allows at most $1,200 a week, for GPT models only. These plans sit alongside the prepaid balance, so you keep one address for Claude Desktop, your OpenAI-format tools and Gemini CLI.