Public API

An open REST API: current language model prices, their change history, budget-based model recommendations, and token/cost calculation. The same data the calculator uses.

✓ no keys or registration✓ JSON (UTF-8)✓ prices per 1M tokens · usd / rub

Base URL

https://api.llmoney.ru/api/v1

Quick start

curl "https://api.llmoney.ru/api/v1/models"

GET/models

— List models

Returns every supported model with current prices. Prices are per 1M tokens in the currency given by the currency field (usd or rub). The tokenizer_type field shows how tokens are counted: tiktoken (local), api_gigachat / api_yandex (via the provider API) or approximation (approximate estimate).

Query parameters

providerstring	Filter by provider slug: openai, anthropic, google, deepseek, meta, mistral, xai, cohere, moonshot, alibaba, zai, minimax, meituan, gigachat, yandex, cloudru and the routers aitunnel, vsegpt, openrouter, vsellm, proxyapi (full list in GET /llms.txt)
free_onlyboolean	Only models with free (local) tokenization. Defaults to false
toolsstring	Only models offering ALL the listed provider-hosted built-in tools (comma-separated): web_search, code_interpreter, file_search, image_generation, computer_use, mcp

Example request

curl "https://api.llmoney.ru/api/v1/models?provider=openai"

Example response

{
  "models": [
    {
      "id": "openai-gpt-5",
      "provider": "openai",
      "provider_display_name": "OpenAI",
      "name": "gpt-5",
      "display_name": "GPT-5",
      "tokenizer_type": "tiktoken",
      "tokenizer_config": { "encoding": "o200k_base" },
      "input_price": 1.25,
      "output_price": 10.0,
      "cached_input_price": 0.125,
      "currency": "usd",
      "is_paid_api": false,
      "supports_byok": true,
      "context_window": 400000,
      "description": null,
      "knowledge_cutoff": "Apr 2025",
      "builtin_tools": ["web_search", "code_interpreter", "file_search", "image_generation", "mcp"]
    }
  ]
}

GET/models/{model_id}

— Model by ID

Returns a single model with its current price. If the model is not found, the response is 404 with a detail field.

Path parameters

model_id*string

Model ID, e.g. openai-gpt-5

Example request

curl "https://api.llmoney.ru/api/v1/models/openai-gpt-5"

Example response

{
  "id": "openai-gpt-5",
  "provider": "openai",
  "provider_display_name": "OpenAI",
  "name": "gpt-5",
  "display_name": "GPT-5",
  "tokenizer_type": "tiktoken",
  "tokenizer_config": { "encoding": "o200k_base" },
  "input_price": 1.25,
  "output_price": 10.0,
  "cached_input_price": 0.125,
  "currency": "usd",
  "is_paid_api": false,
  "supports_byok": true,
  "context_window": 400000,
  "description": null,
  "knowledge_cutoff": "Apr 2025",
  "builtin_tools": ["web_search", "code_interpreter", "file_search", "image_generation", "mcp"]
}

GET/models/{model_id}/price-history

— Model price history

Returns the history of model price changes in chronological order. The source field indicates where the record came from: seed (initial data), manual (manual update) or api (automatic sync).

Path parameters

model_id*string

Model ID, e.g. openai-gpt-5

Example request

curl "https://api.llmoney.ru/api/v1/models/openai-gpt-5/price-history"

Example response

{
  "model_id": "openai-gpt-5",
  "entries": [
    {
      "date": "2025-08-07",
      "input_price": 2.5,
      "output_price": 20.0,
      "cached_input_price": 0.25,
      "source": "seed"
    },
    {
      "date": "2026-07-17",
      "input_price": 1.25,
      "output_price": 10.0,
      "cached_input_price": 0.125,
      "source": "openrouter"
    }
  ]
}

GET/price-changes

— Price change feed

Returns price changes across the entire catalog, newest first: what got cheaper, what got more expensive, by what percentage, and which source the price came from. A record appears only when the price actually changed. The kind values: drop (price decreased), rise (price increased), mixed (input and output moved in opposite directions), new (model added to the catalog).

Query parameters

daysinteger	Window depth in days, 1–1825. Defaults to 90
kindstring	all \| changes (default, without model additions) \| drop \| rise \| new
providerstring	Filter by provider slug
model_idstring	Filter by a single model
min_change_pctnumber	Hide changes smaller than the given percentage. Defaults to 0
limitinteger	Maximum number of records, 1–500. Defaults to 100

Example request

curl "https://api.llmoney.ru/api/v1/price-changes?days=30&kind=drop"

Example response

{
  "summary": {
    "period_days": 30,
    "total": 2,
    "drops": 2,
    "rises": 0,
    "mixed": 0,
    "new_models": 0,
    "models_affected": 2
  },
  "changes": [
    {
      "model_id": "openai-gpt-5",
      "model_name": "GPT-5",
      "provider": "openai",
      "provider_display_name": "OpenAI",
      "currency": "usd",
      "date": "2026-07-17T03:00:00+00:00",
      "kind": "drop",
      "source": "openrouter",
      "input_price": 1.25,
      "output_price": 10.0,
      "cached_input_price": 0.125,
      "previous_input_price": 2.5,
      "previous_output_price": 20.0,
      "previous_cached_input_price": 0.25,
      "input_change_pct": -50.0,
      "output_change_pct": -50.0,
      "cached_change_pct": -50.0
    }
  ]
}

GET/recommend

— Model selection by parameters

Returns a simplified list of models filtered by budget, context window and provider, sorted by the given criterion. The endpoint is designed for programmatic model selection by AI agents. The total_available field is the number of models that passed the filters before limit was applied.

Query parameters

max_input_pricenumber	Maximum input token price per 1M tokens
max_output_pricenumber	Maximum output token price per 1M tokens
min_context_windowinteger	Minimum context window size in tokens
currencystring	Filter by billing currency: usd or rub
providerstring	Filter by provider slug, e.g. openai or gigachat
toolsstring	Only models offering ALL the listed provider-hosted built-in tools (comma-separated): web_search, code_interpreter, file_search, image_generation, computer_use, mcp
sort_bystring	Sort order: input_price (default), output_price or context_window (descending)
limitinteger	Maximum number of models in the response, 1 to 100. Defaults to 10

Example request

curl "https://api.llmoney.ru/api/v1/recommend?max_input_price=5&currency=usd&sort_by=input_price&limit=3"

Example response

{
  "models": [
    {
      "id": "openai-gpt-5-mini",
      "provider": "openai",
      "name": "gpt-5-mini",
      "display_name": "GPT-5 Mini",
      "input_price": 0.25,
      "output_price": 2.0,
      "cached_input_price": 0.025,
      "currency": "usd",
      "context_window": 400000,
      "builtin_tools": ["web_search", "code_interpreter", "file_search", "image_generation", "mcp"]
    }
  ],
  "total_available": 42,
  "filters_applied": {
    "max_input_price": 5.0,
    "currency": "usd",
    "sort_by": "input_price",
    "limit": 3
  }
}

GET/exchange-rate

— USD exchange rate

Returns the current USD/RUB rate (rubles per 1 dollar) that the service uses to convert prices. The source field: cbr (Central Bank of Russia), cbr-mirror (mirror) or fallback (fallback value).

Example request

curl "https://api.llmoney.ru/api/v1/exchange-rate"

Example response

{
  "rate": 78.42,
  "source": "cbr",
  "updated_at": "2026-07-06T11:30:00"
}

POST/calculate

— Token and cost calculation

Counts the number of tokens in the text and the processing cost for each model. Without model_ids the calculation runs over all models with free tokenization. For each model, results returns input_cost, output_cost, total_cost and the tokenizer used (tokenizer_used); the ≈ symbol in it means an approximate count.

Request body (JSON)

text*string	Text to count tokens in
model_idsstring[]	Model IDs. If omitted, all models with free tokenization are used
output_tokensinteger	Expected answer: integer 0–10,000,000. Overrides output_multiplier when supplied; 0 counts input only.
output_multipliernumber	Expected response size relative to the input. Defaults to 1.0
cached_input_rationumber	Share of input tokens served from the prompt cache, 0 to 1. Defaults to 0
toolsobject[]	Tool overhead estimation: an array of objects {type: code_interpreter \| web_search \| file_search \| mcp \| function, count, calls, def_tokens, call_tokens, result_tokens}

Example request

curl -X POST "https://api.llmoney.ru/api/v1/calculate" \
  -H "Content-Type: application/json" \
  -d '{"text": "Привет, мир!", "model_ids": ["openai-gpt-5"]}'

Example response

{
  "input_length": 12,
  "word_count": 2,
  "results": [
    {
      "model_id": "openai-gpt-5",
      "provider": "openai",
      "provider_display_name": "OpenAI",
      "model_name": "gpt-5",
      "display_name": "GPT-5",
      "tokens": 5,
      "cached_tokens": 0,
      "input_cost": 0.00000625,
      "cached_input_cost": 0.0,
      "effective_input_cost": 0.00000625,
      "output_cost": 0.00005,
      "total_cost": 0.00005625,
      "currency": "usd",
      "context_window": 400000,
      "is_available": true,
      "tokenizer_used": "o200k_base",
      "tool_input_tokens": 0,
      "tool_output_tokens": 0,
      "tool_cost": 0.0
    }
  ]
}

POST/tokenize

— Text tokenization

Splits the text into tokens using the chosen tiktoken encoding and returns each token with its ID. Suitable for visualizing tokenization and for exact counts for a specific OpenAI model.

Request body (JSON)

text*string	Text to tokenize
encodingstring	tiktoken encoding: cl100k_base (default), o200k_base, p50k_base or r50k_base

Example request

curl -X POST "https://api.llmoney.ru/api/v1/tokenize" \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello, world!", "encoding": "cl100k_base"}'

Example response

{
  "tokens": [
    { "id": 9906, "text": "Hello" },
    { "id": 11, "text": "," },
    { "id": 1917, "text": " world" },
    { "id": 0, "text": "!" }
  ],
  "total_count": 4,
  "encoding": "cl100k_base"
}

GET/health

— Health check

Service health check. The only endpoint outside the /api/v1 prefix — served from the domain root: https://api.llmoney.ru/health.

Example request

curl "https://api.llmoney.ru/health"

Example response

{
  "status": "healthy"
}

MCP server for AI agents

llmoney-mcp@0.2.0 is available on npm. An agent in Claude Code, Cursor or Windsurf can fetch prices, select models for a budget and estimate request costs. Install Node.js 22+; no LLMoney keys or registration are required.

claude mcp add llmoney -- npx -y llmoney-mcp@0.2.0

For Cursor / Windsurf / Claude Desktop, add the configuration below. On its first run, npx downloads the package and dependencies from npm; access to the private source repository is not required. Text for cost estimation is sent to the LLMoney API; token counts for many models are approximate.

{
  "mcpServers": {
    "llmoney": { "command": "npx", "args": ["-y", "llmoney-mcp@0.2.0"] }
  }
}

list_models— the catalog with current prices, filters, and usd/rub conversion
get_model— prices and capabilities of one model
recommend_model— model recommendations for a budget and context size
estimate_cost— request cost from text or a token count
price_changes— what got cheaper or pricier over a period
price_history— a model's full price history
exchange_rate— USD→RUB rate (Central Bank of Russia)

Private endpoints

The /auth, /api-keys and /call endpoints serve the LLMoney web app itself: OAuth sign-in, storing user API keys (BYOK), and calling models on your behalf. They require authentication and are not part of the public API.

Machine-readable documentation

The full specification is generated automatically and always matches the current API version.

Swagger UI — Interactive documentation — requests can be run right from the browser
ReDoc — The same reference in an alternative format
openapi.json — Machine-readable OpenAPI 3.1 specification
llms.txt — API description for AI agents following the llms.txt convention