diff --git a/service.yaml.example b/service.yaml.example
index 9f73015e4c64ed4e448d0c12ab72de20d63def44..47116f71190c3a65707973a6901982761d157286 100644
--- a/service.yaml.example
+++ b/service.yaml.example
@@ -6,6 +6,9 @@ workflow:
   max_report_generated_num: 1
   recursion_limit: 30
 
+outliner:
+  max_section_num: 2
+
 planner:
   max_step_num: 2
 
diff --git a/src/manager/search_context.py b/src/manager/search_context.py
index 88b017bd5df3c76e08de0124d95070b32ae709ac..eb9ca6baad234ff50957f8d982f31f33bd959270 100644
--- a/src/manager/search_context.py
+++ b/src/manager/search_context.py
@@ -15,6 +15,24 @@ from langgraph.graph import MessagesState
 from pydantic import BaseModel, Field
 
 
+class Section(BaseModel):
+    '''
+    章节模型：表示研究报告大纲结构中的每个章节
+    '''
+    title: str = Field(..., description="章节标题，概括本章节整体目标")
+    description: str = Field(..., description="章节研究任务详细说明，明确指定需要收集的数据或执行的编程任务")
+
+
+class Outline(BaseModel):
+    '''
+    大纲模型：表示研究报告的大纲结构
+    '''
+    language: str = Field(default="zh-CN", description="用户语言：zh-CN、en-US等")
+    thought: str = Field(..., description="研究报告大纲背后的思考过程")
+    title: str = Field(..., description="研究报告标题，概括整体研究任务")
+    sections: List[Section] = Field(default_factory=list, description="最终研究报告的章节")
+
+
 class StepType(str, Enum):
     '''
     任务类型枚举
diff --git a/src/prompts/outliner.md b/src/prompts/outliner.md
new file mode 100644
index 0000000000000000000000000000000000000000..4e6458804cf3fce274c26d7eff83b44271052137
--- /dev/null
+++ b/src/prompts/outliner.md
@@ -0,0 +1,43 @@
+---
+Current Time: {{CURRENT_TIME}}
+---
+
+As a professional Deep Research outliner, skilled in planning systematic research report structures. 
+Your responsibility is to generate a complete outline of the research report based on the given problem, and each item of the outline will be assigned to a team of specialized agents to collect more comprehensive data.
+
+# Core Principles
+- **Comprehensive Coverage**: All aspects + multi-perspective views (mainstream + alternative)
+- **Depth Requirement**: Reject superficial data; require detailed data points + multi-source analysis
+
+## Analysis Framework (8 Dimensions)
+1. **Historical Context**: Evolution timeline  
+2. **Current Status**: Data points + recent developments  
+3. **Future Indicators**: Predictive models + scenario planning  
+4. **Stakeholder Data**: Group impact + perspective mapping  
+5. **Quantitative Data**: Multi-source statistics  
+6. **Qualitative Data**: Case studies + testimonies  
+7. **Comparative Analysis**: Cross-case benchmarking  
+8. **Risk Assessment**: Challenges + contingency plans  
+
+## Execution Constraints
+- Max sections: {{ max_section_num }} (require high focus)
+- Language consistency: **{{ language }}**
+
+## Output Rules
+
+- Keep in mind, directly output the original JSON format of `Outline` without using "```json". 
+- The structure of the `Outline` is defined as follows, and each of the following fields is indispensable.
+
+```ts
+interface Section {
+    title: string; 
+    description: string;  // Each research section
+}
+
+interface Outline {
+    language: string; 
+    thought: string;
+    title: string; 
+    sections: Section[];  // Section list of the final report
+}
+```
diff --git a/src/query_understanding/outliner.py b/src/query_understanding/outliner.py
new file mode 100644
index 0000000000000000000000000000000000000000..f7c89ef45f3434577973284cf08bffb2958c601d
--- /dev/null
+++ b/src/query_understanding/outliner.py
@@ -0,0 +1,77 @@
+# ******************************************************************************
+# Copyright (c) 2025 Huawei Technologies Co., Ltd.
+# jiuwen-deepsearch is licensed under Mulan PSL v2.
+# You can use this software according to the terms and conditions of the Mulan PSL v2.
+# You may obtain a copy of Mulan PSL v2 at:
+#          http://license.coscl.org.cn/MulanPSL2
+# THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
+# EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
+# MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
+# See the Mulan PSL v2 for more details.
+# ******************************************************************************
+
+import json
+import logging
+from json import JSONDecodeError
+
+from langchain_core.exceptions import OutputParserException
+from langchain_core.messages import AIMessage, HumanMessage
+from langchain_core.runnables import RunnableConfig
+
+from src.llm import LLMWrapper
+from src.manager.search_context import SearchContext, Outline
+from src.prompts import apply_system_prompt
+from src.utils.llm_utils import normalize_json_output
+
+logger = logging.getLogger(__name__)
+
+
+class Outliner:
+    def __init__(self):
+        self.llm = LLMWrapper("basic").with_structured_output(schema=Outline, method="json_mode")
+
+    def generate_outline(self, context: SearchContext, config: RunnableConfig) -> dict:
+        """Generating an outline of the report."""
+        logger.info("Outliner starting")
+        messages = apply_system_prompt("outliner", context, config)
+
+        llm_result = ""
+        outline = {}
+        try:
+            # invoke LLM
+            response = self.llm.invoke(messages)
+            llm_result = response.model_dump_json(indent=4)
+            logger.info(f"Outliner LLM result: {llm_result}")
+
+            generated_outline = json.loads(normalize_json_output(llm_result))
+            # validation
+            outline = Outline.model_validate(generated_outline)
+        except JSONDecodeError:
+            logger.error("Outliner LLM response failed JSON deserialization")
+        except OutputParserException as e:
+            logger.error(f"LLM output does not match the structure of the outline: {e}")
+        except Exception as e:
+            logger.error(f"Error when Outliner generating a outline: {e}")
+
+        return {
+            "messages": [AIMessage(name="outliner", content=llm_result)],
+            "current_outline": outline
+        }
+
+
+if __name__ == "__main__":
+    context: SearchContext = {
+        "messages": [
+            HumanMessage("中国平均海拔"),
+        ]
+    }
+
+    config: RunnableConfig = {
+        "configurable": {
+            "max_section_num": 2,
+            "language": "zh-CN"
+        }
+    }
+
+    outliner = Outliner()
+    print(outliner.generate_outline(context, config))