PR-CY API — OpenAI-compatible gateway

/v1/chat/completions over api.pr-cy.ru. Access ships with any paid PR-CY plan: pay in rubles, no foreign cards or VPN, credits are shared across the API and SEO tools.

How to get started

Already on a paid PR-CY plan

Your plan unlocks both the API and SEO tools — they share the same credits balance. The key was created with your account; pick it up in API Settings.

New here

Sign-up is free. API access requires any paid PR-CY plan, payable in rubles without foreign cards or VPN. Credits are shared across the API and SEO tools.

Three steps

If you've used the OpenAI API, the switch takes one line. No new library to install.

1. Grab your API key

Open API Settings in your PR-CY dashboard — the key was created with the account. Just copy it: you send requests, we charge credits.

2. Set base_url

Replace the OpenAI URL with https://api.pr-cy.ru/v1. The key goes in the Authorization: Bearer header.

3. Send the request

POST to /chat/completions with the model and messages fields. If model is omitted, prcy-5.4-mini is used.

Request example

Compose the request in the form — cURL / Python / Node.js code updates automatically. Paid users can hit Send and get a real response below.

model

prcy-5.4-mini

max_tokens (1…32,000)

system (assistant role)

user (what to ask)

cURL

Python

Node.js

updates from the form above

curl https://api.pr-cy.ru/v1/chat/completions \
  -H "Authorization: Bearer $PRCY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "prcy-5.4-mini",
  "max_tokens": 400,
  "messages": [
    {
      "role": "system",
      "content": "You are a PR-CY SEO copywriter."
    },
    {
      "role": "user",
      "content": "Write an SEO title and description for a corner sofas page."
    }
  ]
}'

Response

A standard OpenAI chat.completion object — parses with the same libraries. Example response below.

{
  "id": "chatcmpl-7Qb3f1...",
  "object": "chat.completion",
  "model": "prcy-4o-mini",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Title: Corner sofas — buy affordably with delivery..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 48,
    "completion_tokens": 126,
    "total_tokens": 174
  }
}

Models and credit cost

Pass the identifier in the model field. The gateway currently accepts only GPT-family models. Input — credits per 1000 prompt characters; output — credits per 1000 response tokens.

prcy-4o-minifast

Default basic model: fast, supports images, minimum credits per request.

Context (chars): 128K
Input / 1K chars: 0.5
Output / 1K tokens: 1
Images: yes

prcy-5.4-minifast

GPT-5.4 mini: cheaper than the flagship, good for high-volume work.

Context (chars): 200K
Input / 1K chars: 0.4
Output / 1K tokens: 1
Images: yes

prcy-5.5flagship

Flagship: complex copy, reasoning, long briefs. Available on plans with advanced models.

Context (chars): 200K
Input / 1K chars: 2
Output / 1K tokens: 10
Images: yes

Model	Context (chars)	Max output (tokens)	Input / 1K chars	Output / 1K tokens	Images
Model	Context (chars)	Max output (tokens)	Input / 1K chars	Output / 1K tokens	Images
`prcy-5.4-nano`fast	200K	32K	0.1	0.5	yes
`prcy-5.4-mini`fast	200K	32K	0.4	1	yes
`prcy-4o-mini`fast	128K	16.4K	0.5	1	yes
`prcy-5.5`flagship	200K	32K	2	10	yes
`prcy-5.4`flagship	200K	32K	1	5	yes

Hook it into IDEs and agents

The API is OpenAI-compatible — any client that takes base_url + api_key connects in two lines. Pick a tool, copy the config and drop in your key.

opencode

Continue

Cursor

Aider

~/.config/opencode/opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "prcy": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "PR-CY",
      "options": {
        "baseURL": "https://api.pr-cy.ru/v1",
        "apiKey": "<PRCY_API_KEY>"
      },
      "models": {
        "prcy-4o-mini":  { "name": "PR-CY 4o mini" },
        "prcy-5.4-mini": { "name": "PR-CY 5.4 mini" },
        "prcy-5.5":      { "name": "PR-CY 5.5" }
      }
    }
  }
}

How credits are charged

A single credits balance on your PR-CY plan, for both the API and the SEO tools. Estimate the charge right here.

Prompt = the whole history, not just the last message

All elements of messages are counted: system + the entire conversation. If the total length exceeds the model's context, the overflow is trimmed from the top and isn't billed.

Input in characters — no tokenizer needed

We measure the prompt length with mb_strlen, not in tokens. You can estimate the cost up front without shipping a tokenizer in your code.

Images — billed separately

For vision models, each image_url adds credits on top of the text, by resolution: ⌈width / 512⌉ × ⌈height / 512⌉ × 170 tokens per image.

Two-phase charging

At launch we reserve the prompt cost + 1000 response tokens. After generation we top up if the actual completion_tokens exceeds that. Visible in usage on every response.

Charge calculator

Estimate based on the model's price. Illustrative values.

prcy-5.4-mini

Input characters (prompt)8K

Output tokens (response)2K

5.20 credits per request

input 0.4 per 1K chars · output 1 per 1K tokens · 1 credit ≈ 0.5 ₽

Error codes

Format compatible with OpenAI: HTTP status + JSON with an error.message field.

Code	Type	What it means	What to do
Code	Type	What it means	What to do
`401`	request error	Invalid key	Check the `Authorization: Bearer <key>` header. No spaces or line breaks.
`402`	config or limit	Not enough credits	Credits depleted or the model isn't available on your plan. Top up or pick a smaller model.
`404`	config or limit	Model not found	The `model` field contains an unknown identifier. Check the table above or call `/v1/models`.
`422`	request error	Invalid request	Missing `messages` or malformed body. Verify the JSON and `Content-Type: application/json` header.
`429`	config or limit	Rate limit	Too many requests or credits exhausted on your plan. Back off and retry.
`5xx`	transient	Upstream error	Transient issue at the model provider. Retry with exponential backoff.

FAQ

Is this really compatible with the OpenAI SDK?

Yes. The /v1/chat/completions endpoint mirrors the OpenAI schema: same model, messages, stream, temperature fields. Swap base_url for https://api.pr-cy.ru/v1, plug in the key — the rest of your code stays the same.

Which model runs if I don't specify model?

By default the request goes to prcy-4o-mini — fast and cheap. To pick another, pass its identifier in the model field.

How exactly are credits charged?

From the unified account balance. Input is measured in characters of the prompt (including the system message), output — in tokens of the response. Formula: credits = input × ⌈prompt_chars / 1000⌉ + output × ⌈completion_tokens / 1000⌉. Rounded up, minimum 1 input unit. The actual volume comes back in the usage field on every response. 1 credit ≈ 0.5 ₽ — exact price depends on the plan.

Which model has the lowest charge?

The cheapest are the nano- and mini-versions: prcy-5.4-nano, prcy-4o-mini, prcy-5.4-mini. Flagships prcy-5.5 and prcy-5.4 cost more but handle complex tasks better.

Are streaming and images supported?

Streaming via "stream": true (Server-Sent Events). Models marked Yes in the Images column accept images — as image_url blocks in content, just like OpenAI vision.

Will other models be available — Claude, Gemini, DeepSeek?

Yes, we're testing. The gateway currently routes only the GPT family; the rest we use in internal tools (chat, AI tools). As we verify compatibility with the OpenAI schema, we'll open them through the same /v1/chat/completions. Identifiers and prices will land in the table above.

Where do I get the API key?

In your PR-CY dashboard, under API Settings. The key was created with your account — just copy it. Keep it in an environment variable, don't publish it in client-side code.

Do I need a separate plan for the API?

No. The API uses the same credits balance as the SEO tools. Buy any PR-CY plan — both the API and the tools work off the same balance.

🍪 By using this website, you agree to the processing of cookies and collection of technical data to improve website performance in accordance with our privacy policy.