openai · rm-openai · Jun 16, 2025 · Jun 16, 2025
diff --git a/examples/basic/prompt_template.py b/examples/basic/prompt_template.py
@@ -0,0 +1,79 @@
+import argparse
+import asyncio
+import random
+
+from agents import Agent, GenerateDynamicPromptData, Runner
+
+"""
+NOTE: This example will not work out of the box, because the default prompt ID will not be available
+in your project.
+
+To use it, please:
+1. Go to https://platform.openai.com/playground/prompts
+2. Create a new prompt variable, `poem_style`.
+3. Create a system prompt with the content:
+```
+Write a poem in {{poem_style}}
+```
+4. Run the example with the `--prompt-id` flag.
+"""
+
+DEFAULT_PROMPT_ID = "pmpt_6850729e8ba481939fd439e058c69ee004afaa19c520b78b"
+
+
+class DynamicContext:
+    def __init__(self, prompt_id: str):
+        self.prompt_id = prompt_id
+        self.poem_style = random.choice(["limerick", "haiku", "ballad"])
+        print(f"[debug] DynamicContext initialized with poem_style: {self.poem_style}")
+
+
+async def _get_dynamic_prompt(data: GenerateDynamicPromptData):
+    ctx: DynamicContext = data.context.context
+    return {
+        "id": ctx.prompt_id,
+        "version": "1",
+        "variables": {
+            "poem_style": ctx.poem_style,
+        },
+    }
+
+
+async def dynamic_prompt(prompt_id: str):
+    context = DynamicContext(prompt_id)
+
+    agent = Agent(
+        name="Assistant",
+        prompt=_get_dynamic_prompt,
+    )
+
+    result = await Runner.run(agent, "Tell me about recursion in programming.", context=context)
+    print(result.final_output)
+
+
+async def static_prompt(prompt_id: str):
+    agent = Agent(
+        name="Assistant",
+        prompt={
+            "id": prompt_id,
+            "version": "1",
+            "variables": {
+                "poem_style": "limerick",
+            },
+        },
+    )
+
+    result = await Runner.run(agent, "Tell me about recursion in programming.")
+    print(result.final_output)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--dynamic", action="store_true")
+    parser.add_argument("--prompt-id", type=str, default=DEFAULT_PROMPT_ID)
+    args = parser.parse_args()
+
+    if args.dynamic:
+        asyncio.run(dynamic_prompt(args.prompt_id))
+    else:
+        asyncio.run(static_prompt(args.prompt_id))
diff --git a/pyproject.toml b/pyproject.toml
@@ -7,7 +7,7 @@ requires-python = ">=3.9"
 license = "MIT"
 authors = [{ name = "OpenAI", email = "[email protected]" }]
 dependencies = [
-    "openai>=1.81.0",
+    "openai>=1.87.0",
     "pydantic>=2.10, <3",
     "griffe>=1.5.6, <2",
     "typing-extensions>=4.12.2, <5",

diff --git a/src/agents/__init__.py b/src/agents/__init__.py
@@ -45,6 +45,7 @@
 from .models.openai_chatcompletions import OpenAIChatCompletionsModel
 from .models.openai_provider import OpenAIProvider
 from .models.openai_responses import OpenAIResponsesModel
+from .prompts import DynamicPromptFunction, GenerateDynamicPromptData, Prompt
 from .repl import run_demo_loop
 from .result import RunResult, RunResultStreaming
 from .run import RunConfig, Runner
@@ -178,6 +179,9 @@ def enable_verbose_stdout_logging():
     "AgentsException",
     "InputGuardrailTripwireTriggered",
     "OutputGuardrailTripwireTriggered",
+    "DynamicPromptFunction",
+    "GenerateDynamicPromptData",
+    "Prompt",
     "MaxTurnsExceeded",
     "ModelBehaviorError",
     "UserError",

diff --git a/src/agents/agent.py b/src/agents/agent.py
@@ -7,6 +7,7 @@
 from dataclasses import dataclass, field
 from typing import TYPE_CHECKING, Any, Callable, Generic, Literal, cast
 
+from openai.types.responses.response_prompt_param import ResponsePromptParam
 from typing_extensions import NotRequired, TypeAlias, TypedDict
 
 from .agent_output import AgentOutputSchemaBase
@@ -17,6 +18,7 @@
 from .mcp import MCPUtil
 from .model_settings import ModelSettings
 from .models.interface import Model
+from .prompts import DynamicPromptFunction, Prompt, PromptUtil
 from .run_context import RunContextWrapper, TContext
 from .tool import FunctionTool, FunctionToolResult, Tool, function_tool
 from .util import _transforms
@@ -95,6 +97,12 @@ class Agent(Generic[TContext]):
     return a string.
     """
 
+    prompt: Prompt | DynamicPromptFunction | None = None
+    """A prompt object (or a function that returns a Prompt). Prompts allow you to dynamically
+    configure the instructions, tools and other config for an agent outside of your code. Only
+    usable with OpenAI models, using the Responses API.
+    """
+
     handoff_description: str | None = None
     """A description of the agent. This is used when the agent is used as a handoff, so that an
     LLM knows what it does and when to invoke it.
@@ -242,6 +250,12 @@ async def get_system_prompt(self, run_context: RunContextWrapper[TContext]) -> s
 
         return None
 
+    async def get_prompt(
+        self, run_context: RunContextWrapper[TContext]
+    ) -> ResponsePromptParam | None:
+        """Get the prompt for the agent."""
+        return await PromptUtil.to_model_input(self.prompt, run_context, self)
+
     async def get_mcp_tools(self) -> list[Tool]:
         """Fetches the available tools from the MCP servers."""
         convert_schemas_to_strict = self.mcp_config.get("convert_schemas_to_strict", False)

diff --git a/src/agents/extensions/models/litellm_model.py b/src/agents/extensions/models/litellm_model.py
@@ -71,6 +71,7 @@ async def get_response(
         handoffs: list[Handoff],
         tracing: ModelTracing,
         previous_response_id: str | None,
+        prompt: Any | None = None,
     ) -> ModelResponse:
         with generation_span(
             model=str(self.model),
@@ -88,6 +89,7 @@ async def get_response(
                 span_generation,
                 tracing,
                 stream=False,
+                prompt=prompt,
             )
 
             assert isinstance(response.choices[0], litellm.types.utils.Choices)
@@ -153,8 +155,8 @@ async def stream_response(
         output_schema: AgentOutputSchemaBase | None,
         handoffs: list[Handoff],
         tracing: ModelTracing,
-        *,
         previous_response_id: str | None,
+        prompt: Any | None = None,
     ) -> AsyncIterator[TResponseStreamEvent]:
         with generation_span(
             model=str(self.model),
@@ -172,6 +174,7 @@ async def stream_response(
                 span_generation,
                 tracing,
                 stream=True,
+                prompt=prompt,
             )
 
             final_response: Response | None = None
@@ -202,6 +205,7 @@ async def _fetch_response(
         span: Span[GenerationSpanData],
         tracing: ModelTracing,
         stream: Literal[True],
+        prompt: Any | None = None,
     ) -> tuple[Response, AsyncStream[ChatCompletionChunk]]: ...
 
     @overload
@@ -216,6 +220,7 @@ async def _fetch_response(
         span: Span[GenerationSpanData],
         tracing: ModelTracing,
         stream: Literal[False],
+        prompt: Any | None = None,
     ) -> litellm.types.utils.ModelResponse: ...
 
     async def _fetch_response(
@@ -229,6 +234,7 @@ async def _fetch_response(
         span: Span[GenerationSpanData],
         tracing: ModelTracing,
         stream: bool = False,
+        prompt: Any | None = None,
     ) -> litellm.types.utils.ModelResponse | tuple[Response, AsyncStream[ChatCompletionChunk]]:
         converted_messages = Converter.items_to_messages(input)
 

diff --git a/src/agents/models/interface.py b/src/agents/models/interface.py
@@ -5,6 +5,8 @@
 from collections.abc import AsyncIterator
 from typing import TYPE_CHECKING
 
+from openai.types.responses.response_prompt_param import ResponsePromptParam
+
 from ..agent_output import AgentOutputSchemaBase
 from ..handoffs import Handoff
 from ..items import ModelResponse, TResponseInputItem, TResponseStreamEvent
@@ -46,6 +48,7 @@ async def get_response(
         tracing: ModelTracing,
         *,
         previous_response_id: str | None,
+        prompt: ResponsePromptParam | None,
     ) -> ModelResponse:
         """Get a response from the model.
 
@@ -59,6 +62,7 @@ async def get_response(
             tracing: Tracing configuration.
             previous_response_id: the ID of the previous response. Generally not used by the model,
                 except for the OpenAI Responses API.
+            prompt: The prompt config to use for the model.
 
         Returns:
             The full model response.
@@ -77,6 +81,7 @@ def stream_response(
         tracing: ModelTracing,
         *,
         previous_response_id: str | None,
+        prompt: ResponsePromptParam | None,
     ) -> AsyncIterator[TResponseStreamEvent]:
         """Stream a response from the model.
 
@@ -90,6 +95,7 @@ def stream_response(
             tracing: Tracing configuration.
             previous_response_id: the ID of the previous response. Generally not used by the model,
                 except for the OpenAI Responses API.
+            prompt: The prompt config to use for the model.
 
         Returns:
             An iterator of response stream events, in OpenAI Responses format.

diff --git a/src/agents/models/openai_chatcompletions.py b/src/agents/models/openai_chatcompletions.py
@@ -9,6 +9,7 @@
 from openai.types import ChatModel
 from openai.types.chat import ChatCompletion, ChatCompletionChunk
 from openai.types.responses import Response
+from openai.types.responses.response_prompt_param import ResponsePromptParam
 from openai.types.responses.response_usage import InputTokensDetails, OutputTokensDetails
 
 from .. import _debug
@@ -53,6 +54,7 @@ async def get_response(
         handoffs: list[Handoff],
         tracing: ModelTracing,
         previous_response_id: str | None,
+        prompt: ResponsePromptParam | None = None,
     ) -> ModelResponse:
         with generation_span(
             model=str(self.model),
@@ -69,6 +71,7 @@ async def get_response(
                 span_generation,
                 tracing,
                 stream=False,
+                prompt=prompt,
             )
 
             first_choice = response.choices[0]
@@ -136,8 +139,8 @@ async def stream_response(
         output_schema: AgentOutputSchemaBase | None,
         handoffs: list[Handoff],
         tracing: ModelTracing,
-        *,
         previous_response_id: str | None,
+        prompt: ResponsePromptParam | None = None,
     ) -> AsyncIterator[TResponseStreamEvent]:
         """
         Yields a partial message as it is generated, as well as the usage information.
@@ -157,6 +160,7 @@ async def stream_response(
                 span_generation,
                 tracing,
                 stream=True,
+                prompt=prompt,
             )
 
             final_response: Response | None = None
@@ -187,6 +191,7 @@ async def _fetch_response(
         span: Span[GenerationSpanData],
         tracing: ModelTracing,
         stream: Literal[True],
+        prompt: ResponsePromptParam | None = None,
     ) -> tuple[Response, AsyncStream[ChatCompletionChunk]]: ...
 
     @overload
@@ -201,6 +206,7 @@ async def _fetch_response(
         span: Span[GenerationSpanData],
         tracing: ModelTracing,
         stream: Literal[False],
+        prompt: ResponsePromptParam | None = None,
     ) -> ChatCompletion: ...
 
     async def _fetch_response(
@@ -214,6 +220,7 @@ async def _fetch_response(
         span: Span[GenerationSpanData],
         tracing: ModelTracing,
         stream: bool = False,
+        prompt: ResponsePromptParam | None = None,
     ) -> ChatCompletion | tuple[Response, AsyncStream[ChatCompletionChunk]]:
         converted_messages = Converter.items_to_messages(input)
 

diff --git a/src/agents/models/openai_responses.py b/src/agents/models/openai_responses.py
@@ -17,6 +17,7 @@
     WebSearchToolParam,
     response_create_params,
 )
+from openai.types.responses.response_prompt_param import ResponsePromptParam
 
 from .. import _debug
 from ..agent_output import AgentOutputSchemaBase
@@ -74,6 +75,7 @@ async def get_response(
         handoffs: list[Handoff],
         tracing: ModelTracing,
         previous_response_id: str | None,
+        prompt: ResponsePromptParam | None = None,
     ) -> ModelResponse:
         with response_span(disabled=tracing.is_disabled()) as span_response:
             try:
@@ -86,6 +88,7 @@ async def get_response(
                     handoffs,
                     previous_response_id,
                     stream=False,
+                    prompt=prompt,
                 )
 
                 if _debug.DONT_LOG_MODEL_DATA:
@@ -141,6 +144,7 @@ async def stream_response(
         handoffs: list[Handoff],
         tracing: ModelTracing,
         previous_response_id: str | None,
+        prompt: ResponsePromptParam | None = None,
     ) -> AsyncIterator[ResponseStreamEvent]:
         """
         Yields a partial message as it is generated, as well as the usage information.
@@ -156,6 +160,7 @@ async def stream_response(
                     handoffs,
                     previous_response_id,
                     stream=True,
+                    prompt=prompt,
                 )
 
                 final_response: Response | None = None
@@ -192,6 +197,7 @@ async def _fetch_response(
         handoffs: list[Handoff],
         previous_response_id: str | None,
         stream: Literal[True],
+        prompt: ResponsePromptParam | None = None,
     ) -> AsyncStream[ResponseStreamEvent]: ...
 
     @overload
@@ -205,6 +211,7 @@ async def _fetch_response(
         handoffs: list[Handoff],
         previous_response_id: str | None,
         stream: Literal[False],
+        prompt: ResponsePromptParam | None = None,
     ) -> Response: ...
 
     async def _fetch_response(
@@ -217,6 +224,7 @@ async def _fetch_response(
         handoffs: list[Handoff],
         previous_response_id: str | None,
         stream: Literal[True] | Literal[False] = False,
+        prompt: ResponsePromptParam | None = None,
     ) -> Response | AsyncStream[ResponseStreamEvent]:
         list_input = ItemHelpers.input_to_new_input_list(input)
 
@@ -252,6 +260,7 @@ async def _fetch_response(
             input=list_input,
             include=converted_tools.includes,
             tools=converted_tools.tools,
+            prompt=self._non_null_or_not_given(prompt),
             temperature=self._non_null_or_not_given(model_settings.temperature),
             top_p=self._non_null_or_not_given(model_settings.top_p),
             truncation=self._non_null_or_not_given(model_settings.truncation),