Claude Code with any LLM (₹ pricing)
Run Claude Code against Qwen, Llama, DeepSeek using unoblox gateway. One base URL, one key, a GST invoice at each top-up. No international card.
Use Claude Code with any model, billed in ₹
Claude Code (the agentic coding CLI) talks to models via the /v1/messages Anthropic-compatible endpoint. Point it at unoblox, swap three settings, and run Qwen, DeepSeek, or Llama at ₹-native rates on a GST invoice — while keeping Claude itself available for the calls where it's worth the premium.
Setup in 3 steps
- Get your API key from unoblox.ai/sign-in. Copy the
ub-gw-…value. - In Claude Code's settings, change:
- API endpoint:
https://api.unoblox.ai/v1 - API key: paste your
ub-gw-…key - Model: pick
qwen/qwen3-235b-a22bordeepseek-ai/deepseek-v3.2
- API endpoint:
- Verify: run any command, then check the line item in
/dashboard.
Why unoblox for Claude Code
| Metric | unoblox | Typical foreign SaaS |
|---|---|---|
| Billing | GST tax invoice issued at each top-up | Pay-per-use, card only |
| Model choice | OpenRouter-class catalog | Often locked to one vendor |
| Currency | ₹ native billing | Foreign-currency card required |
| Latency | Asia-based gateway | International routing |
Model pricing (per 1M input/output tokens)
- Qwen3.8-27B ₹0 / ₹0 (vision, fast)
- DeepSeek V3.2 ₹26.32 / ₹38.47 (reasoning, cheap)
- Claude Sonnet ₹202.46 / ₹1,012.28 (benchmark quality)
- Qwen3.8-27B ₹0 (free tier, reasoning)
When to pick unoblox for Claude Code
You're a good fit if you're using Claude Code to:
- Build with open-source models (Qwen, Llama, DeepSeek) instead of only Claude
- Avoid foreign-card billing friction on a recurring developer tool
- Claim GST input tax credit on your AI tooling spend
- Get a GST tax invoice issued at each top-up across every model, instead of separate vendor bills
You don't have to pick one model forever — keep Claude wired up for the hardest refactors and route routine edits, boilerplate, and test-writing to a cheaper open model, all through the same /v1/messages endpoint.
Frequently asked questions
Q: Will my scripts break?
No. Claude Code and your local agentic runners use the same /v1/messages spec. Swap URL and key, test one invocation, run.
Q: Which models does Claude Code support through unoblox?
Any model on unoblox's /v1/messages endpoint — Qwen, DeepSeek, Claude, GPT, and Llama variants. Pick per task based on the speed/cost tradeoff you need.
Q: How do I switch back to Claude's own API?
Reverse the two settings: endpoint back to api.anthropic.com, key to your Anthropic key, model to your usual Claude model. Takes about a minute.
Q: Do I get an invoice? Yes. Every call generates a line with a GST tax invoice issued at each top-up, and you can claim input tax credit on it. Export invoices anytime from the dashboard.
Q: How fast is the gateway? Latency is dominated by the model you choose, not the network — unoblox serves from Asia-based infrastructure with no international hop added on top.
Q: Can I use tool use / function calling?
Yes. /v1/messages supports tools, JSON mode, and streaming — no API changes needed beyond the base URL and key.
Get started in rupees → https://unoblox.ai/sign-in
Start building in rupees
Call every major model through one OpenAI-compatible endpoint, billed in ₹ on a GST invoice.