Skip to main content

​
Short answer

The Base URL must match the protocol expected by your tool. Claude Code usually uses the Anthropic-compatible root address. Codex CLI, Cursor, Cline, OpenCode, OpenClaw, and other external tools usually use the OpenAI-compatible /v1 address. In LLMEasy:
Use caseBase URL
Claude Code / Anthropic-compatiblehttps://www.llmeasy.ru
Codex CLI / OpenAI-compatiblehttps://www.llmeasy.ru/v1
Cursor, Cline, OpenCode, OpenClawhttps://www.llmeasy.ru/v1
Do not guess the path. API Key, Base URL, and Model ID must come from the same provider or gateway.

​
Why there are different Base URLs

Different SDKs append paths in different ways.
Protocol or tool typeCommon behavior
OpenAI-compatible SDKUsually calls resources such as chat completions or responses under /v1
Anthropic-compatible SDKUsually appends message endpoint paths itself, so you should not add /v1 again
Built-in tool providerMay append paths automatically, or may ask you for a complete Base URL
The same model name does not mean the same Base URL works everywhere. The real match is between tool protocol, Base URL, API Key, Key group, and Model ID.

​
Quick reference

ToolRecommended setup
Claude Codehttps://www.llmeasy.ru, with a Claude group API Key
Claude Desktop GatewayFollow the Gateway page, including the auth scheme
Codex CLIhttps://www.llmeasy.ru/v1, with a GPT group API Key
Cursorhttps://www.llmeasy.ru/v1, with a GPT group API Key
Clinehttps://www.llmeasy.ru/v1, using an OpenAI Compatible provider
OpenCode / OpenClawhttps://www.llmeasy.ru/v1, with a GPT group API Key

​
Common errors

Error or symptomPossible cause
404 Not FoundThe Base URL path does not match the SDK’s expected path
model not foundThe Model ID does not belong to the current Key group or provider
invalid_api_keyThe API Key and Base URL are from different services
Authentication works, but the model is unavailableThe Key group and Model ID do not match
Requests keep failingAn Anthropic-compatible address was used in an OpenAI-compatible tool, or the reverse
  1. Confirm whether the tool expects OpenAI-compatible or Anthropic-compatible API.
  2. Check whether the Base URL matches that protocol.
  3. Check whether the API Key is from the same service.
  4. Check whether the Model ID comes from the model plaza and belongs to the current Key group.
  5. Send a minimal test request, such as hello.

​
Code examples

OpenAI-compatible tools usually use the /v1 address: 
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://www.llmeasy.ru/v1"
)

response = client.chat.completions.create(
    model="your-model-id",
    messages=[{"role": "user", "content": "hello"}]
)
Claude / Anthropic-compatible setups usually use the root address: 
import anthropic

client = anthropic.Anthropic(
    api_key="YOUR_API_KEY",
    base_url="https://www.llmeasy.ru"
)

message = client.messages.create(
    model="your-claude-model-id",
    max_tokens=1024,
    messages=[{"role": "user", "content": "hello"}]
)

​
About LLMEasy

LLMEasy provides both Anthropic-compatible and OpenAI-compatible access. You can manage API Keys, balance, and usage records in the same dashboard, but each client still needs the protocol address it supports. Quick rule: Claude Code uses https://www.llmeasy.ru; Codex CLI, Cursor, Cline, OpenCode, and OpenClaw use https://www.llmeasy.ru/v1.