Langflow in Russia: Claude, GPT, DeepSeek and Qwen for agents and RAG without a VPN
Langflow is an open-source visual builder (MIT license) for assembling AI agents, chatbots and RAG over your own documents from blocks. It has no models of its own — you need a provider key. Add the OpenAI Compatible provider once in settings with the address https://www.ruapi.ai/v1 and a RuAPI key, and Claude, GPT, Gemini, DeepSeek and Qwen appear in the Language Model, Agent and Embedding Model components. No VPN, pay in USDT.
# Profile → Settings → Model Providers → OpenAI Compatible
Base URL: https://www.ruapi.ai/v1
API Key: sk-...The OpenAI Compatible tab is the provider setup in Langflow 1.11 and newer. The Anthropic tab is for the Anthropic component with Claude’s native API. The Terminal tab installs via Python; Langflow Desktop and the langflowai/langflow Docker image are also available.
What gets in the way of connecting models to Langflow from Russia
Without a provider key, flows that use AI will not run.
Agents, chatbots and RAG in Langflow call the model through a provider component, and it needs a key. Anthropic and OpenAI do not offer their APIs in Russia: it is missing from their supported-country lists (anthropic.com/supported-countries, developers.openai.com, checked 10 Nov 2026). A RuAPI key is issued right after you sign up with email, Google, Yandex or Telegram.
One key and one bill per provider.
If one flow runs its agent on Claude and computes embeddings with an OpenAI model, you usually need two accounts and two payments. Through RuAPI every model works with one key and one balance, and the console log shows what each request cost.
The Langflow server is often in Russia.
Langflow is installed on your own server, and model requests go straight from it. Keeping a VPN on a server is awkward and unreliable. www.ruapi.ai is reachable from Russia directly, so there is nothing to set up on the server.
How to connect: 3 steps
Works for Langflow Desktop, Docker and a Python install.
Sign up for RuAPI and top up
Sign up at www.ruapi.ai with email, Google, Yandex or Telegram. The minimum top-up is 5 USDT on TRC20, BEP20, TON or Polygon; USDC is accepted too. How to buy USDT for rubles via SBP is shown at docs.ruapi.ai/en/buy-usdt. Foreign bank cards also work.
A new account usually gets $0.05 right away for deepseek-flash, claude-haiku-5-5 and gpt-6-luna — enough to test your first chat flow. Embeddings (document search) unlock after your first top-up.
Create a key and add the provider in Langflow
Console → “API keys” → “Create key”. In Langflow open your profile → Settings → Model Providers → OpenAI Compatible, enter Base URL https://www.ruapi.ai/v1 and the API Key, and click Save. Langflow loads the list of models your key can use: tick the ones you need under Language Models and Embedding Models. To keep the list short, allow only the models you need for the key in the RuAPI console.
Pick models in your components
- In a Language Model or Agent component choose the OpenAI Compatible provider and a model, for example claude-sonnet-5-5 or deepseek-flash.
- In Embedding Model choose text-embedding-3-small or text-embedding-3-large. Langflow shows every model in both lists, but a chat model will not work in the embeddings field — double-check your choice.
- For Claude’s native API use the Anthropic component: Anthropic API URL — https://www.ruapi.ai; type the model name by hand (the list refresh button calls api.anthropic.com).
- Keep the key in Settings → Global Variables as a Credential variable and pick it with the globe icon next to the key field.
RuAPI addresses and keys: docs.ruapi.ai/en/quickstart · tool calling: docs.ruapi.ai/en/function-calling
What to enter in Langflow
OpenAI Compatible
Base URL: https://www.ruapi.ai/v1
API Key: sk-…
Anthropic component
Anthropic API URL: https://www.ruapi.ai
Model Name: claude-sonnet-5-5
Embeddings
text-embedding-3-small — $0.024 per 1M tokens
text-embedding-3-large
Prices of models for Langflow
In US dollars per 1M tokens, as on the pricing page. No subscription: only the tokens your requests use are deducted.
| model | input $/M | output $/M |
|---|---|---|
| claude-sonnet-5-5 | $2.40 cache read $0.12 | $12.00 |
| claude-haiku-5-5 | $0.12 cache read $0.0120 >100K: $0.60 | $0.60 >100K: $3.00 |
| gpt-6.1-sol | $2.40 cache read $0.12 >272K: $4.80 | $12.00 >272K: $18.00 |
| gpt-6-luna | $0.12 cache read $0.0120 >272K: $0.24 | $0.60 >272K: $0.90 |
| gemini-4-argon | $2.40 cache read $0.12 | $12.00 |
| deepseek-v4-pro | $1.58 cache read $0.0600 | $4.75 |
| deepseek-flash | $0.17 cache read $0.0120 | $1.44 |
| qwen3.8-flash | $0.14 cache read $0.0288 | $0.49 |
Input is the price of new input tokens (prompt, history, retrieved document chunks), output of the model’s reply. By our estimate, a RAG request with 5K tokens in and a 500-token reply costs about $0.002 on deepseek-flash and about $0.018 on claude-sonnet-5-5. Indexing 500K tokens of documents with text-embedding-3-small costs about $0.012. Your account’s prices may be lower.
The grey “>100K” row is the vendor’s long-prompt rate: if the prompt is over the threshold, the whole request is billed at it.