Overview
TypedDict definitions for beta message creation request parameters, including the base parameter class and its streaming/non-streaming variants.
Description
This module defines the MessageCreateParams TypedDict hierarchy for the beta Messages API. The core class MessageCreateParamsBase declares all fields available when creating a beta message, including required fields (max_tokens, messages, model) and numerous optional fields for controlling model behavior, tool use, thinking, citations, containers, context management, MCP servers, and more.
Two variant classes extend the base: MessageCreateParamsNonStreaming sets stream to Literal[False] (optional), while MessageCreateParamsStreaming requires stream: Literal[True]. The union MessageCreateParams = Union[MessageCreateParamsNonStreaming, MessageCreateParamsStreaming] provides the full type for the beta messages.create() method. A generic variant ParseMessageCreateParamsBase adds an output_format field typed as type[ResponseFormatT] for structured output parsing.
This is the most comprehensive parameter definition in the SDK, as the beta Messages API supports the full range of experimental features including extended thinking, tool use (client and server tools), web search, web fetch, code execution, MCP toolsets, containers, context management, and structured output formats. The file is auto-generated from the OpenAPI specification by the Stainless code generation tool.
Usage
Use these types when constructing request parameters for the beta messages.create() method, or when type-annotating functions that accept or produce beta message creation parameters. The types ensure compile-time type safety and provide IDE autocompletion for all available parameters.
Code Reference
Source Location
- Repository: Anthropic SDK Python
- File:
src/anthropic/types/beta/message_create_params.py (354 lines)
Classes
| Class |
Parent |
Description
|
MessageCreateParamsBase |
TypedDict |
Base parameter definition with all fields (total=False except Required fields)
|
ParseMessageCreateParamsBase |
MessageCreateParamsBase, Generic[ResponseFormatT] |
Generic variant with typed output_format for structured output parsing
|
MessageCreateParamsNonStreaming |
MessageCreateParamsBase |
Non-streaming variant (stream: Literal[False])
|
MessageCreateParamsStreaming |
MessageCreateParamsBase |
Streaming variant (stream: Required[Literal[True]])
|
OutputFormat |
TypedDict |
JSON schema output format definition
|
Type Aliases
Container: TypeAlias = Union[BetaContainerParams, str]
MessageCreateParams = Union[MessageCreateParamsNonStreaming, MessageCreateParamsStreaming]
Signature
class MessageCreateParamsBase(TypedDict, total=False):
max_tokens: Required[int]
messages: Required[Iterable[BetaMessageParam]]
model: Required[ModelParam]
container: Optional[Container]
context_management: Optional[BetaContextManagementConfigParam]
inference_geo: Optional[str]
mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam]
metadata: BetaMetadataParam
output_config: BetaOutputConfigParam
output_format: Optional[BetaJSONOutputFormatParam] # Deprecated
service_tier: Literal["auto", "standard_only"]
speed: Optional[Literal["standard", "fast"]]
stop_sequences: SequenceNotStr[str]
system: Union[str, Iterable[BetaTextBlockParam]]
temperature: float
thinking: BetaThinkingConfigParam
tool_choice: BetaToolChoiceParam
tools: Iterable[BetaToolUnionParam]
top_k: int
top_p: float
betas: Annotated[List[AnthropicBetaParam], PropertyInfo(alias="anthropic-beta")]
class MessageCreateParamsNonStreaming(MessageCreateParamsBase, total=False):
stream: Literal[False]
class MessageCreateParamsStreaming(MessageCreateParamsBase):
stream: Required[Literal[True]]
Import
from anthropic.types.beta import MessageCreateParams
from anthropic.types.beta.message_create_params import (
MessageCreateParamsBase,
MessageCreateParamsNonStreaming,
MessageCreateParamsStreaming,
)
I/O Contract
Required Fields
| Field |
Type |
Description
|
max_tokens |
int |
Maximum tokens to generate before stopping. Limits vary by model.
|
messages |
Iterable[BetaMessageParam] |
Input conversation messages. Alternating user/assistant turns. Limit of 100,000 messages.
|
model |
ModelParam |
Model identifier (e.g., "claude-sonnet-4-20250514")
|
Optional Fields -- Model Behavior
| Field |
Type |
Description
|
temperature |
float |
Randomness (0.0 to 1.0, default 1.0). Lower for analytical, higher for creative.
|
top_k |
int |
Top-K sampling. Advanced use only.
|
top_p |
float |
Nucleus sampling. Do not combine with temperature.
|
stop_sequences |
SequenceNotStr[str] |
Custom text sequences that trigger generation stop.
|
system |
Union[str, Iterable[BetaTextBlockParam]] |
System prompt providing context and instructions.
|
speed |
None |
Inference speed mode; "fast" enables high throughput.
|
service_tier |
Literal["auto", "standard_only"] |
Priority vs standard capacity selection.
|
Optional Fields -- Tools and Features
| Field |
Type |
Description
|
tools |
Iterable[BetaToolUnionParam] |
Tool definitions (client tools, server tools, computer use, text editor, bash, web search, etc.)
|
tool_choice |
BetaToolChoiceParam |
How the model should use tools (any, auto, none, specific tool)
|
thinking |
BetaThinkingConfigParam |
Extended thinking configuration. Minimum 1,024 token budget.
|
mcp_servers |
Iterable[BetaRequestMCPServerURLDefinitionParam] |
MCP server definitions for remote tool access
|
container |
None |
Container identifier for code execution persistence
|
context_management |
None |
Context management configuration for multi-turn optimization
|
Optional Fields -- Output and Metadata
| Field |
Type |
Description
|
metadata |
BetaMetadataParam |
Request metadata object
|
output_config |
BetaOutputConfigParam |
Output configuration (format, etc.)
|
output_format |
None |
Deprecated: Use output_config.format instead. JSON schema output format.
|
inference_geo |
None |
Geographic region for inference processing
|
betas |
List[AnthropicBetaParam] |
Beta version flags (sent as anthropic-beta header)
|
Streaming Variants
| Variant |
stream Field |
Usage
|
MessageCreateParamsNonStreaming |
Literal[False] (optional) |
Standard request returning a complete BetaMessage
|
MessageCreateParamsStreaming |
Required[Literal[True]] |
SSE streaming request returning incremental events
|
Usage Examples
from anthropic.types.beta.message_create_params import (
MessageCreateParamsBase,
MessageCreateParamsNonStreaming,
MessageCreateParamsStreaming,
)
# Non-streaming params
params: MessageCreateParamsNonStreaming = {
"model": "claude-sonnet-4-20250514",
"max_tokens": 1024,
"messages": [
{"role": "user", "content": "What is the capital of France?"}
],
}
# Streaming params
stream_params: MessageCreateParamsStreaming = {
"model": "claude-sonnet-4-20250514",
"max_tokens": 1024,
"messages": [
{"role": "user", "content": "Tell me a story."}
],
"stream": True,
}
# With extended thinking
thinking_params: MessageCreateParamsNonStreaming = {
"model": "claude-sonnet-4-20250514",
"max_tokens": 16000,
"messages": [
{"role": "user", "content": "Solve this complex math problem..."}
],
"thinking": {
"type": "enabled",
"budget_tokens": 10000,
},
}
# With tools
tool_params: MessageCreateParamsNonStreaming = {
"model": "claude-sonnet-4-20250514",
"max_tokens": 1024,
"messages": [
{"role": "user", "content": "What is the weather in Tokyo?"}
],
"tools": [
{
"name": "get_weather",
"description": "Get weather for a city",
"input_schema": {
"type": "object",
"properties": {"city": {"type": "string"}},
"required": ["city"],
},
}
],
"tool_choice": {"type": "auto"},
}
# With system prompt and temperature
full_params: MessageCreateParamsNonStreaming = {
"model": "claude-sonnet-4-20250514",
"max_tokens": 2048,
"system": "You are a helpful assistant that responds in haiku form.",
"messages": [
{"role": "user", "content": "Tell me about the ocean."}
],
"temperature": 0.7,
}
Related Pages
Implements Principle
Part Of
Uses Types
See Also
