diff --git a/backend-fastapi/ai_platform/nodes/builtin/template_node.py b/backend-fastapi/ai_platform/nodes/builtin/template_node.py
index 873efce..220a55d 100644
--- a/backend-fastapi/ai_platform/nodes/builtin/template_node.py
+++ b/backend-fastapi/ai_platform/nodes/builtin/template_node.py
@@ -2,8 +2,11 @@
 模板渲染节点
 """
 import logging
+import time
 from typing import Any, Dict
 
+from sqlalchemy import select
+
 from ..base import BaseNode, NodeContext, NodeResult
 from ..registry import NodeRegistry
 
@@ -61,6 +64,144 @@ class TemplateNode(BaseNode):
                 success=False,
                 error=str(e),
             )
+
+    async def execute_async(self, context: NodeContext) -> NodeResult:
+        """配置 agent_code 时，以智能体身份执行当前模板任务。"""
+        agent_code = (self.config.get('agent_code') or '').strip()
+        if not agent_code:
+            return self.execute(context)
+
+        start_time = time.time()
+        try:
+            if not context.db_session:
+                return NodeResult(success=False, error='智能体模板节点需要数据库会话')
+
+            from ai_platform.models import Agent
+            from ai_platform.services.llm_service import LLMService
+
+            result = await context.db_session.execute(
+                select(Agent).where(
+                    Agent.code == agent_code,
+                    Agent.is_deleted == False,
+                    Agent.status == 'published',
+                )
+            )
+            agent = result.scalar_one_or_none()
+            if not agent:
+                return NodeResult(success=False, error=f'智能体不存在或未发布: {agent_code}')
+
+            model_id = self.config.get('model_id') or agent.model_id
+            if not model_id:
+                return NodeResult(success=False, error=f'智能体未配置模型: {agent.name}({agent.code})')
+
+            template = self.config.get('template', '')
+            rendered_prompt = context.resolve_template(template)
+            if not rendered_prompt.strip():
+                return NodeResult(success=False, error='智能体模板节点缺少任务内容')
+
+            messages = [
+                {'role': 'system', 'content': self._build_agent_system_prompt(agent)},
+                {'role': 'user', 'content': self._build_agent_user_prompt(context, rendered_prompt)},
+            ]
+            response = await LLMService(context.db_session).chat_async(
+                model_id=str(model_id),
+                messages=messages,
+                temperature=self.config.get(
+                    'temperature',
+                    agent.temperature if agent.temperature is not None else 0.7,
+                ),
+                max_tokens=self.config.get('max_tokens', agent.max_tokens or 2048),
+            )
+
+            elapsed_time = int((time.time() - start_time) * 1000)
+            output_variable = self.config.get('output_variable') or f'{agent.code}_result'
+            output_variables = {
+                output_variable: response.content,
+                f'{output_variable}_agent_code': agent.code,
+                f'{output_variable}_agent_name': agent.name,
+                f'{output_variable}_tokens': response.total_tokens,
+            }
+
+            return NodeResult(
+                success=True,
+                output=response.content,
+                output_variables=output_variables,
+                tokens_used=response.total_tokens,
+                elapsed_time=elapsed_time,
+                metadata={
+                    'agent_code': agent.code,
+                    'agent_name': agent.name,
+                    'model_id': str(model_id),
+                    'model': response.model,
+                    'prompt_tokens': response.prompt_tokens,
+                    'completion_tokens': response.completion_tokens,
+                    'output_variable': output_variable,
+                    'frontend_output_variables': output_variables,
+                },
+            )
+
+        except Exception as e:
+            logger.exception(f'智能体模板节点执行失败: {e}')
+            return NodeResult(
+                success=False,
+                error=str(e),
+                elapsed_time=int((time.time() - start_time) * 1000),
+            )
+
+    @staticmethod
+    def _build_agent_system_prompt(agent: Any) -> str:
+        base_prompt = agent.system_prompt or ''
+        persona = agent.persona or {}
+
+        if not base_prompt and isinstance(persona, dict):
+            parts = []
+            if persona.get('role'):
+                parts.append(persona['role'])
+            if persona.get('skills'):
+                parts.append('你擅长：' + '、'.join(str(item) for item in persona['skills']) + '。')
+            if persona.get('constraints'):
+                parts.append('注意事项：\n' + '\n'.join(f'- {item}' for item in persona['constraints']))
+            if persona.get('background'):
+                parts.append(str(persona['background']))
+            base_prompt = '\n\n'.join(parts)
+
+        if not base_prompt:
+            base_prompt = '你是一个智能体，请用中文完成当前任务。'
+
+        return (
+            f'{base_prompt}\n\n'
+            '你正在作为流程编排中的协作智能体执行当前步骤。'
+            '请只输出本步骤的结论、交付证据、风险和下一步，避免空泛说明。'
+        )
+
+    @staticmethod
+    def _format_context_value(value: Any, max_length: int = 1200) -> str:
+        if isinstance(value, str):
+            text = value
+        else:
+            import json
+            try:
+                text = json.dumps(value, ensure_ascii=False, indent=2)
+            except TypeError:
+                text = str(value)
+        return text if len(text) <= max_length else f'{text[:max_length]}...'
+
+    def _build_agent_user_prompt(self, context: NodeContext, rendered_prompt: str) -> str:
+        context_lines = []
+        for key, value in context.variables.items():
+            if key.startswith('_') or key.endswith('_tokens') or key.endswith('_agent_code') or key.endswith('_agent_name'):
+                continue
+            context_lines.append(f'- {key}: {self._format_context_value(value)}')
+            if len('\n'.join(context_lines)) > 6000:
+                context_lines.append('- 其余上下文因长度限制已省略')
+                break
+
+        sections = [f'当前步骤任务：\n{rendered_prompt}']
+        if context.previous_output:
+            sections.append(f'上一节点输出：\n{self._format_context_value(context.previous_output, 2000)}')
+        if context_lines:
+            sections.append('可用工作流上下文：\n' + '\n'.join(context_lines))
+        return '\n\n'.join(sections)
     
     @classmethod
     def get_config_schema(cls) -> Dict[str, Any]:
@@ -79,6 +220,16 @@ class TemplateNode(BaseNode):
                     'title': '输出变量名',
                     'default': 'template_result',
                 },
+                'agent_code': {
+                    'type': 'string',
+                    'title': '智能体编码',
+                    'description': '配置后会以该智能体身份调用模型执行模板任务',
+                },
+                'model_id': {
+                    'type': 'string',
+                    'title': '覆盖模型 ID',
+                    'description': '为空时使用智能体默认模型',
+                },
             },
             'required': ['template'],
         }
diff --git a/backend-fastapi/ai_platform/services/workflow_service.py b/backend-fastapi/ai_platform/services/workflow_service.py
index 5494065..635f4ec 100644
--- a/backend-fastapi/ai_platform/services/workflow_service.py
+++ b/backend-fastapi/ai_platform/services/workflow_service.py
@@ -64,6 +64,10 @@ def _make_execution_log_entry(
     }
 
 
+def _node_result_metadata(result: NodeResult) -> dict:
+    return dict(result.metadata or {})
+
+
 def _snapshot_node_inputs(context: NodeContext) -> dict:
     return {
         'variables': dict(context.variables),
@@ -535,6 +539,7 @@ class AIWorkflowService:
                         node_type,
                         status='completed',
                         output=result.output,
+                        metadata=_node_result_metadata(result),
                         inputs=copy.deepcopy(node_inputs),
                     ))
                 break
@@ -563,6 +568,7 @@ class AIWorkflowService:
                 error=result.error,
                 elapsed_time=elapsed,
                 tokens_used=result.tokens_used,
+                metadata=_node_result_metadata(result),
                 inputs=copy.deepcopy(node_inputs),
             ))
             
@@ -725,6 +731,7 @@ class AIWorkflowService:
                     'error': result.error,
                     'elapsed_time': elapsed,
                     'tokens_used': result.tokens_used,
+                    'metadata': _node_result_metadata(result),
                     'branch': branch_start_id,
                 })
                 
@@ -940,6 +947,7 @@ class AIWorkflowService:
                     'error': result.error,
                     'elapsed_time': elapsed,
                     'tokens_used': result.tokens_used,
+                    'metadata': _node_result_metadata(result),
                     'branch': branch_id,
                 })
                 
@@ -1228,6 +1236,7 @@ class AIWorkflowService:
                     'error': result.error,
                     'elapsed_time': elapsed,
                     'tokens_used': result.tokens_used,
+                    'metadata': _node_result_metadata(result),
                     'loop_iteration': iteration,
                 })
                 
@@ -1733,6 +1742,7 @@ class AIWorkflowService:
                 error=result.error,
                 elapsed_time=elapsed,
                 tokens_used=result.tokens_used,
+                metadata=_node_result_metadata(result),
                 inputs=copy.deepcopy(node_inputs),
             )
             logs.append(log_entry)
@@ -2201,6 +2211,7 @@ class AIWorkflowService:
                     'status': 'waiting',
                     'output': result.output,
                     'elapsed_time': elapsed,
+                    'metadata': _node_result_metadata(result),
                 }
                 logs.append(log_entry)
                 
@@ -2235,6 +2246,7 @@ class AIWorkflowService:
                 'error': result.error,
                 'elapsed_time': elapsed,
                 'tokens_used': result.tokens_used,
+                'metadata': _node_result_metadata(result),
             }
             logs.append(log_entry)
             
diff --git a/backend-fastapi/scripts/seed_multica_org_agents.py b/backend-fastapi/scripts/seed_multica_org_agents.py
index 25739e8..615ebe9 100644
--- a/backend-fastapi/scripts/seed_multica_org_agents.py
+++ b/backend-fastapi/scripts/seed_multica_org_agents.py
@@ -53,21 +53,12 @@ def validate_fixture(fixture: Dict[str, Any]) -> Dict[str, int]:
     edges = definition.get("edges", [])
     agent_codes = {agent.get("code") for agent in agents}
     workflow_agent_codes = {
-        node.get("agent_code")
+        (node.get("data") or {}).get("agent_code")
         for node in nodes
-        if isinstance(node, dict) and node.get("agent_code")
+        if isinstance(node, dict) and (node.get("data") or {}).get("agent_code")
     }
     node_types = {node.get("type") for node in nodes}
 
-    required_agent_codes = {
-        "multica_product_manager",
-        "business_requirements_analyst",
-        "system_architect",
-        "frontend_engineer",
-        "backend_engineer",
-        "qa_engineer",
-        "project_manager",
-    }
     required_persona_fields = {
         "role",
         "skills",
@@ -75,10 +66,8 @@ def validate_fixture(fixture: Dict[str, Any]) -> Dict[str, int]:
         "background",
         "examples",
     }
-    required_node_types = {"start", "end", "condition", "template", "parallel", "merge"}
+    required_node_types = {"start", "end", "template", "parallel", "merge"}
 
-    missing_agents = sorted(required_agent_codes - agent_codes)
-    extra_agents = sorted(agent_codes - required_agent_codes)
     missing_workflow_agents = sorted(workflow_agent_codes - agent_codes)
     missing_persona_fields = {
         agent.get("code"): sorted(required_persona_fields - set((agent.get("persona") or {}).keys()))
@@ -92,17 +81,13 @@ def validate_fixture(fixture: Dict[str, Any]) -> Dict[str, int]:
     )
     project_manager_workflow = (project_manager or {}).get("workflow_code")
     if (
-        missing_agents
-        or extra_agents
-        or missing_workflow_agents
+        missing_workflow_agents
         or missing_persona_fields
         or missing_node_types
         or project_manager_workflow != "multica_org_collaboration_flow"
     ):
         raise ValueError(
-            f"Fixture validation failed: missing_agents={missing_agents}, "
-            f"extra_agents={extra_agents}, "
-            f"missing_workflow_agents={missing_workflow_agents}, "
+            f"Fixture validation failed: missing_workflow_agents={missing_workflow_agents}, "
             f"missing_persona_fields={missing_persona_fields}, "
             f"missing_node_types={missing_node_types}, "
             f"project_manager_workflow={project_manager_workflow}"
@@ -116,6 +101,24 @@ def validate_fixture(fixture: Dict[str, Any]) -> Dict[str, int]:
     }
 
 
+async def resolve_default_chat_model_id(session) -> str | None:
+    from sqlalchemy import select
+
+    from ai_platform.models.model import LLMModel
+
+    result = await session.execute(
+        select(LLMModel)
+        .where(
+            LLMModel.is_deleted == False,
+            LLMModel.is_active == True,
+            LLMModel.model_type == "chat",
+        )
+        .order_by(LLMModel.sort.desc(), LLMModel.sys_create_datetime.desc())
+    )
+    model = result.scalars().first()
+    return str(model.id) if model else None
+
+
 async def upsert_workflow(session, workflow_payload: Dict[str, Any]) -> AIWorkflow:
     from sqlalchemy import select
 
@@ -166,8 +169,15 @@ async def upsert_agents(session, fixture: Dict[str, Any], workflow_id: str) -> i
     from app.base_model import generate_nanoid
 
     count = 0
+    default_chat_model_id = await resolve_default_chat_model_id(session)
     for agent_fixture in fixture["agents"]:
         payload = build_agent_payload(agent_fixture, workflow_id)
+        if (
+            payload.get("mode", "autonomous") == "autonomous"
+            and not payload.get("model_id")
+            and default_chat_model_id
+        ):
+            payload["model_id"] = default_chat_model_id
         result = await session.execute(select(Agent).where(Agent.code == payload["code"]))
         agent = result.scalar_one_or_none()
         if agent:
diff --git a/web/apps/web-ele/src/views/ai-platform/workflow-runs/modules/detail-dialog.vue b/web/apps/web-ele/src/views/ai-platform/workflow-runs/modules/detail-dialog.vue
index 10d9baf..49837de 100644
--- a/web/apps/web-ele/src/views/ai-platform/workflow-runs/modules/detail-dialog.vue
+++ b/web/apps/web-ele/src/views/ai-platform/workflow-runs/modules/detail-dialog.vue
@@ -148,6 +148,10 @@ function previewValue(value: any) {
   }
 }
 
+function getLogMeta(log: any) {
+  return log?.metadata || {};
+}
+
 function selectLog(nodeId?: string) {
   if (!nodeId) return;
   activeNodeId.value = nodeId;
@@ -287,6 +291,34 @@ defineExpose({ open });
                     <div v-if="log.error" class="text-destructive break-words">
                       {{ log.error }}
                     </div>
+                    <div
+                      v-if="getLogMeta(log).agent_code"
+                      class="border-border bg-muted/40 grid grid-cols-2 gap-2 rounded-md border p-2"
+                    >
+                      <div>
+                        <div class="text-muted-foreground">智能体</div>
+                        <div class="font-medium">
+                          {{ getLogMeta(log).agent_name || getLogMeta(log).agent_code }}
+                        </div>
+                      </div>
+                      <div>
+                        <div class="text-muted-foreground">编码</div>
+                        <div class="font-mono">{{ getLogMeta(log).agent_code }}</div>
+                      </div>
+                      <div>
+                        <div class="text-muted-foreground">模型</div>
+                        <div class="break-words">
+                          {{ getLogMeta(log).model || getLogMeta(log).model_id || '-' }}
+                        </div>
+                      </div>
+                      <div>
+                        <div class="text-muted-foreground">Prompt / Completion</div>
+                        <div>
+                          {{ getLogMeta(log).prompt_tokens || 0 }} /
+                          {{ getLogMeta(log).completion_tokens || 0 }}
+                        </div>
+                      </div>
+                    </div>
                     <div>
                       <div class="text-muted-foreground mb-1 font-medium">
                         {{ $t('ai-platform.workflowRuns.detail.inputsOutputs') }} - 输入