### Switching to Multi-Agent Mode in Coze Source: https://www.coze.cn/open/docs/guides/multiagent This guide outlines the steps to transition an AI agent from single-agent mode to multi-agent mode within the Coze development platform. It details navigating the workspace, selecting an agent, and activating the multi-agent configuration. ```text 1. Log in to the Coze development platform. 2. Select 'Workspace' from the left navigation bar and choose the target workspace from the space list at the top of the page. 3. On the Develop page, select 'Agent'. 4. Click on the target agent or create a new one in the selected space. 5. Select 'Multi-Agents Mode' on the agent's orchestration page. ``` -------------------------------- ### Coze Multi-Agent Mode Example: Real-time Translation Agent Source: https://www.coze.cn/open/docs/guides/multiagent This example demonstrates building a multi-agent translation system in Coze. It involves creating a 'Real-time Translation' agent, configuring a parent agent for task distribution, and setting up three child agents for Chinese, Japanese, and Korean translation. ```text 1. Create an agent named 'Real-time Translation'. 2. Switch to Multi-Agent Mode and input the global prompt. 3. Configure the parent Agent with an 'Applicable Scenario' like 'Translate user input to the target language'. 4. Connect and configure three child Agent nodes for Chinese, Japanese, and Korean translation. For the 'Translate to Chinese' agent: - Rename the Agent to 'Translate to Chinese'. - Set 'Applicable Scenario' to 'Translate user input to Chinese'. - Input the Agent Prompt detailing the translation task. 5. Copy the 'Translate to Chinese' agent twice, renaming them for Japanese and Korean translation, and update their respective 'Applicable Scenarios' and prompts. 6. Test the translation tasks in the preview and debug area. ``` -------------------------------- ### Creating a Multi-Agent Mode Agent in Coze Source: https://www.coze.cn/open/docs/guides/multiagent This section details the process of building an AI agent using the Multi-Agent Mode in Coze. It covers initial agent creation, mode switching, global configuration, adding and connecting nodes, and debugging the agent's functionality. ```text Step 1: Create Agent and Set Mode 1. Log in to the Coze development platform. 2. Click ⊕ in the upper left corner of the page. 3. Enter the agent name and functional description, then click the generate icon next to the icon to automatically generate an avatar. 4. On the agent's orchestration page, click 'Single Agent Mode' and select 'Multi-Agents Mode'. Step 2: Configure Global Settings In the agent's orchestration panel, describe the agent's persona and add other configurations as needed. These settings are global and apply to all added agents. Step 3: Add Nodes Add and connect nodes in the central canvas area. The 'Start' node is the default starting point. 'Agent' nodes represent individual agents that can perform tasks. 'Sub-Agent' nodes allow integrating pre-published agents. 'Global Jump Condition' nodes enable conditional routing. Step 4: Debug Agent Test the agent in the preview and debug area. You can also click the run icon on an agent to debug a specific node. ``` -------------------------------- ### Coze Agent Node Configuration Source: https://www.coze.cn/open/docs/guides/multiagent This describes the configuration options for an 'Agent' node within Coze's Multi-Agent Mode. It includes renaming, copying, model settings, defining applicable scenarios, and setting agent prompts and skills. ```text Agent Node Configuration: - Click the settings icon (three dots) to change Agent settings. - Rename the Agent for clarity. - Create a copy to duplicate the Agent. - Configure the Large Language Model (LLM) used by the Agent. - Define 'Applicable Scenarios' to guide task distribution. - Input 'Agent Prompt' to define the Agent's operational logic and task execution steps. - Add 'Skills' (tools, workflows, knowledge bases) using the '+' button. - Configure 'User Question Suggestions' to generate follow-up questions. ``` -------------------------------- ### Coze Sub-Agent Node Configuration Source: https://www.coze.cn/open/docs/guides/multiagent Details the configuration for 'Sub-Agent' nodes in Coze's Multi-Agent Mode, which allow integrating existing single-agent agents. It covers viewing details, copying, defining applicable scenarios, and managing user question suggestions. ```text Sub-Agent Node Configuration: - Click the settings icon (three dots) to change Agent settings. - Create a copy to duplicate the Sub-Agent. - Click 'Agent Details' to view and update the agent configuration. - Define 'Applicable Scenarios' to guide task distribution. - Manage 'User Question Suggestions', either following the original agent's state or custom settings. ``` -------------------------------- ### Coze Agent Mode Switching Configuration Retention Source: https://www.coze.cn/open/docs/guides/multiagent This table explains how agent configurations are retained when switching between single-agent and multi-agent modes in Coze. It details which settings persist and which are reset during these transitions. ```text Mode Switching Effects: From Single-Agent to Multi-Agent Mode: - Retained: Persona and response logic, triggers, variables, databases, opening statements, quick commands, background images, and voice. - Quick commands default to automatic node assignment but can be manually assigned. - If it's the first switch to Multi-Agent Mode, added workflows, plugin tools, and knowledge bases are retained and added to the first default agent. From Multi-Agent to Single-Agent Mode: - Retained: Persona and response logic, triggers, variables, databases, opening statements, background images, and voice. - Reset: Quick commands, workflows, plugin tools, and knowledge bases added to each agent are not retained. ``` === COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.