Kumo vs OpenRouter
The OpenRouter alternative that just works from Russia
Same GPT, Claude and Gemini, same OpenAI-compatible API — paid with a Russian card or SBP, reachable without a VPN, with no prompt logging. Swap the base URL: the code stays yours and the question of how to pay goes away
What matters
Four differences that decide it
Both gateways stand in front of the same models. The difference is how you get in, who serves the model to you, and what is left behind after the call
Pay from Russia — no workarounds
A Russian bank card or SBP, a balance in rubles. No foreign card, no intermediaries, no crypto — top up and get to work
Reachable without a VPN
The endpoint answers from Russia directly. Your server, your laptop and your CI reach the models the same way they reach any other API
You pick the supplier
One model is often served by several suppliers — faster, cheaper or with a bigger context window. The catalog shows each one’s price, and you decide before the call rather than finding out on the invoice
Prompts are not logged
Request and response bodies are not stored. Only billing metadata remains — model, tokens, time — exactly what your finance team needs
Axis by axis
How Kumo differs from OpenRouter
Rules rather than prices: they do not change because a lab revised a rate. Today’s rates are in the model catalog
| What is being compared | Kumo | OpenRouter |
|---|---|---|
| Payment from Russia | Russian cards and SBP, balance in rubles | Russian cards are not accepted — a foreign card is required |
| Account currency | Rubles or dollars — the rate is shown before you top up | USD only |
| How the price is set | Every model’s price at every supplier is in the catalog before you call | The provider’s list rate passed through as-is |
| Suppliers of one model | Several per model — you choose who serves it | By default the gateway picks the supplier |
| Works without a VPN | Yes — the endpoint is reachable from Russia | Access from Russia is not guaranteed |
| API compatibility | OpenAI-compatible — swap one base URL, keep your SDK | OpenAI-compatible — swap one base URL, keep your SDK |
| Prompt content | Not logged — only billing metadata is kept | Depends on settings — see OpenRouter’s privacy policy |
| Spend controls | A cap on every key, one balance across all models, spend in one console | Per-key spend limits |
| Model catalog | Curated flagships from OpenAI, Anthropic, Google, xAI and DeepSeek — plus models on request | A very wide catalog, including niche and community models |
Terms on either side move. Verify OpenRouter’s fees and rules on openrouter.ai and Kumo’s current rates in the model catalog
Migration
Moving off OpenRouter is one diff
Kumo speaks the same OpenAI-compatible protocol OpenRouter does. Your SDK, streaming, tool calls and structured output keep working — the base URL and the key change, and the model id comes from the Kumo catalog
from openai import OpenAI client = OpenAI(removed line base_url="https://openrouter.ai/api/v1",added line base_url="https://api.kumorouter.com/v1",removed line api_key=os.environ["OPENROUTER_API_KEY"],added line api_key=os.environ["KUMO_API_KEY"],)resp = client.chat.completions.create( model="anthropic/claude-opus-4-8", messages=[{"role": "user", "content": "Hello!"}],)The base URL and the key change; the model id is the catalog’s, and if it differs from the one you use, the diff shows it. Streaming, tools and structured output behave the same
Decision guide
An honest read on when each wins
Use Kumo when…
- You pay from Russia — Russian cards and SBP, a ruble balance, no VPN and no workarounds
- You want to pick the supplier of a model yourself and see the price before the call, not after
- Your team needs one balance with per-key caps — finance reconciles one report instead of five invoices
- Privacy is a security-review requirement: prompt bodies are not logged
- The model you need is not in the catalog — for teams with volume we add models on request
OpenRouter still wins when…
- You need the long tail — hundreds of niche and community models in one catalog
- You pay comfortably with a foreign card
- Public model rankings and a large community around the service matter to you
Still unsure? Run the same workload for a week on both. Both are drop-in — only the base URL changes — so the experiment costs one line
FAQ
Questions people ask before switching
The six asked most often by teams moving off OpenRouter. The rest is in the general FAQ and on the blog
Run your next request through Kumo
One base URL, a ruble balance, the price shown before the call. Or talk to us about volume for your team