Skip to main content

Claude Code Native

The claude_code_native provider is a specialized variant of Anthropic designed for Claude Code, Anthropic's official CLI for Claude. This provider uses Bearer authentication and automatically includes a specific system prompt optimized for CLI workflows.

Key Differences from Anthropic​

FeatureAnthropicsClaude Code Native
Authenticationx-api-key headerAuthorization: Bearer <token> header
Required Headersanthropic-versionanthropic-version, anthropic-beta: oauth-2025-04-20
System PromptUser-definedAlways prepends: "You are Claude Code, Anthropic's official CLI for Claude."
API Endpointhttps://api.anthropic.com/https://api.anthropic.com/ (same endpoint, different auth/headers)
PropertyDetails
DescriptionSpecialized Anthropic provider for Claude Code CLI with OAuth authentication and system prompt
Provider Route on LiteLLMclaude_code_native/ (add this prefix to model names)
Provider DocClaude Code ↗
API Endpointhttps://api.anthropic.com
Supported Endpoints/chat/completions, /v1/messages (passthrough)

Supported Models​

Not all Anthropic models are supported through the claude_code_native provider. The supported models are

  • claude-sonnet-4-5
  • claude-opus-4-5
  • claude-haiku-4-5

Supported OpenAI Parameters​

The claude_code_native provider supports all Anthropic parameters:

"stream",
"stop",
"temperature",
"top_p",
"max_tokens",
"max_completion_tokens",
"tools",
"tool_choice",
"extra_headers",
"parallel_tool_calls",
"response_format",
"user",
"reasoning_effort",
info

Notes:

  • The provider automatically adds the required system prompt for Claude Code
  • All system messages provided by the user are appended after the Claude Code system prompt
  • The provider requires an OAuth bearer token (not an API key)
  • max_tokens defaults to 4096 if not specified

API Keys and Authentication​

The claude_code_native provider uses OAuth Bearer authentication and requires the CLAUDE_CODE_API_KEY environment variable:

import os

os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"  # Used as Bearer token
# os.environ["CLAUDE_CODE_API_BASE"] = ""  # [OPTIONAL] Custom API base
note

The provider automatically converts the API key to a Bearer token by setting the Authorization: Bearer <token> header instead of x-api-key: <token>.

info

Priority for API Key:

  1. Explicit api_key parameter in completion calls takes highest priority
  2. The CLAUDE_CODE_API_KEY environment variable is required if no explicit api_key is provided

Usage​

Basic Usage​

import os
from litellm import completion

# Set your OAuth bearer token
os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

messages = [{"role": "user", "content": "Hey! how's it going?"}]
response = completion(model="claude_code_native/claude-opus-4-20250514", messages=messages)
print(response)

Usage - Streaming​

import os
from litellm import completion

# Set your OAuth bearer token
os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

messages = [{"role": "user", "content": "Hey! how's it going?"}]
response = completion(model="claude_code_native/claude-opus-4-20250514", messages=messages, stream=True)
for chunk in response:
    print(chunk["choices"][0]["delta"]["content"])  # same as openai format

Usage with Custom System Message​

When you provide your own system message, it will be appended after the Claude Code system prompt:

import os
from litellm import completion

os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

messages = [
    {"role": "system", "content": "You are a helpful coding assistant."},
    {"role": "user", "content": "Help me debug this code."}
]
response = completion(model="claude_code_native/claude-3-5-sonnet-20240620", messages=messages)
print(response)

The actual system messages sent to Anthropic will be:

  1. "You are Claude Code, Anthropic's official CLI for Claude." (auto-added)
  2. "You are a helpful coding assistant." (user-provided)

Usage with LiteLLM Proxy​

1. Save key in your environment​

export CLAUDE_CODE_API_KEY="your-oauth-bearer-token"

2. Start the proxy​

model_list:
  - model_name: claude-code-sonnet-4 ### RECEIVED MODEL NAME ###
    litellm_params: # all params accepted by litellm.completion()
      model: claude_code_native/claude-sonnet-4-20250514 ### MODEL NAME sent to `litellm.completion()` ###
      api_key: "os.environ/CLAUDE_CODE_API_KEY" # does os.getenv("CLAUDE_CODE_API_KEY")
litellm --config /path/to/config.yaml

3. Test it​

curl --location 'http://0.0.0.0:4000/chat/completions' \
--header 'Content-Type: application/json' \
--data ' {
      "model": "claude-code-sonnet-4",
      "messages": [
        {
          "role": "user",
          "content": "what llm are you"
        }
      ]
    }
'

Supported Models​

Model NameFunction CallOAuth Required
claude-sonnet-4-5completion('claude_code_native/claude-sonnet-4-5-20250929', messages)os.environ['CLAUDE_CODE_API_KEY']
claude-opus-4completion('claude_code_native/claude-opus-4-20250514', messages)os.environ['CLAUDE_CODE_API_KEY']
claude-sonnet-4completion('claude_code_native/claude-sonnet-4-20250514', messages)os.environ['CLAUDE_CODE_API_KEY']
claude-3.7completion('claude_code_native/claude-3-7-sonnet-20250219', messages)os.environ['CLAUDE_CODE_API_KEY']
claude-3-5-sonnetcompletion('claude_code_native/claude-3-5-sonnet-20240620', messages)os.environ['CLAUDE_CODE_API_KEY']
claude-3-haikucompletion('claude_code_native/claude-3-haiku-20240307', messages)os.environ['CLAUDE_CODE_API_KEY']
claude-3-opuscompletion('claude_code_native/claude-3-opus-20240229', messages)os.environ['CLAUDE_CODE_API_KEY']
claude-3-sonnetcompletion('claude_code_native/claude-3-sonnet-20240229', messages)os.environ['CLAUDE_CODE_API_KEY']
claude-2.1completion('claude_code_native/claude-2.1', messages)os.environ['CLAUDE_CODE_API_KEY']
claude-2completion('claude_code_native/claude-2', messages)os.environ['CLAUDE_CODE_API_KEY']

Function/Tool Calling​

The claude_code_native provider supports all Anthropic tool calling features:

from litellm import completion
import os

# Set OAuth bearer token
os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

tools = [
    {
        "type": "function",
        "function": {
            "name": "get_current_weather",
            "description": "Get the current weather in a given location",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {
                        "type": "string",
                        "description": "The city and state, e.g. San Francisco, CA",
                    },
                    "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
                },
                "required": ["location"],
            },
        },
    }
]
messages = [{"role": "user", "content": "What's the weather like in Boston today?"}]

response = completion(
    model="claude_code_native/claude-3-opus-20240229",
    messages=messages,
    tools=tools,
    tool_choice="auto",
)
print(response)

Prompt Caching​

The claude_code_native provider supports Anthropic's prompt caching feature. See the Anthropic documentation for detailed examples with the cache_control parameter.

response = await litellm.acompletion(
    model="claude_code_native/claude-3-5-sonnet-20240620",
    messages=[
        {
            "role": "system",
            "content": [
                {
                    "type": "text",
                    "text": "You are an AI assistant tasked with analyzing legal documents.",
                },
                {
                    "type": "text",
                    "text": "Here is the full text of a complex legal agreement",
                    "cache_control": {"type": "ephemeral"},
                },
            ],
        },
        {
            "role": "user",
            "content": "what are the key terms and conditions in this agreement?",
        },
    ]
)

Structured Outputs​

The claude_code_native provider supports Anthropic's structured outputs feature. See the Anthropic documentation for details.

from litellm import completion

response = completion(
    model="claude_code_native/claude-sonnet-4-5-20250929",
    messages=[{"role": "user", "content": "What is the capital of France?"}],
    response_format={
        "type": "json_schema",
        "json_schema": {
            "name": "capital_response",
            "strict": True,
            "schema": {
                "type": "object",
                "properties": {
                    "country": {"type": "string"},
                    "capital": {"type": "string"}
                },
                "required": ["country", "capital"],
                "additionalProperties": False
            }
        }
    }
)

print(response.choices[0].message.content)

Vision​

All Anthropic vision capabilities are supported by claude_code_native:

from litellm import completion
import os
import base64

os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

def encode_image(image_path):
    with open(image_path, "rb") as image_file:
        return base64.b64encode(image_file.read()).decode("utf-8")

image_path = "../proxy/cached_logo.jpg"
base64_image = encode_image(image_path)

resp = litellm.completion(
    model="claude_code_native/claude-3-opus-20240229",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Whats in this image?"},
                {
                    "type": "image_url",
                    "image_url": {
                        "url": "data:image/jpeg;base64," + base64_image
                    },
                },
            ],
        }
    ],
)
print(f"\nResponse: {resp}")

MCP Tool Calling​

MCP tool calling is supported with the claude_code_native provider:

import os 
from litellm import completion

os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

tools=[
    {
        "type": "mcp",
        "server_label": "deepwiki",
        "server_url": "https://mcp.deepwiki.com/mcp",
        "require_approval": "never",
    },
]

response = completion(
    model="claude_code_native/claude-sonnet-4-20250514",
    messages=[{"role": "user", "content": "Who won the World Cup in 2022?"}],
    tools=tools
)

Hosted Tools​

The claude_code_native provider supports Anthropic's hosted tools (Computer, Text Editor, Web Search, Memory):

Computer Use​

from litellm import completion
import os

os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

tools = [
    {
        "type": "computer_20241022",
        "function": {
            "name": "computer",
            "parameters": {
                "display_height_px": 100,
                "display_width_px": 100,
                "display_number": 1,
            },
        },
    }
]
model = "claude_code_native/claude-3-5-sonnet-20241022"
messages = [{"role": "user", "content": "Save a picture of a cat to my desktop."}]

resp = completion(
    model=model,
    messages=messages,
    tools=tools,
)
print(resp)
from litellm import completion
import os

os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

model = "claude_code_native/claude-3-5-sonnet-20241022"
messages = [{"role": "user", "content": "What's the weather like today?"}]

resp = completion(
    model=model,
    messages=messages,
    web_search_options={
        "search_context_size": "medium",
        "user_location": {
            "type": "approximate",
            "approximate": {
                "city": "San Francisco",
            },
        }
    }
)
print(resp)

Memory​

from litellm import completion
import os

os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

tools = [{
    "type": "memory_20250818",
    "name": "memory"
}]

model = "claude_code_native/claude-sonnet-4-5-20250929" 
messages = [{"role": "user", "content": "Please remember that my favorite color is blue."}]

response = completion(
    model=model,
    messages=messages,
    tools=tools,
)
print(response)

Passing Extra Headers​

from litellm import completion
import os

os.environ["CLAUDE_CODE_API_KEY"] = "your-oauth-bearer-token"

messages = [{"role": "user", "content": "What is Claude Code Native?"}]
response = completion(
    model="claude_code_native/claude-3-5-sonnet-20240620", 
    messages=messages, 
    extra_headers={"anthropic-beta": "max-tokens-3-5-sonnet-2024-07-15"}
)
info

The provider automatically merges any extra headers with its required headers (anthropic-beta: oauth-2025-04-20 and anthropic-version: 2023-06-01). The provider's required headers take precedence.

Comparison with Anthropic Provider​

Featureanthropic/claude_code_native/
Authenticationx-api-key headerAuthorization: Bearer header
System PromptUser-provided onlyAuto: "You are Claude Code, Anthropic's official CLI for Claude." + User-provided
Required Headersanthropic-versionanthropic-version, anthropic-beta: oauth-2025-04-20
Use CaseGeneral purposeClaude Code CLI applications
Tool SupportFullFull (same as Anthropic)
Prompt CachingSupportedSupported
VisionSupportedSupported

API Reference​

For detailed API information, see:

note

The claude_code_native provider builds on top of the Anthropic provider implementation. All Anthropic-specific features and parameters are fully supported.