Custom Agents
Create specialized AI assistants with custom instructions and knowledge
Custom Agents
Custom Agents are specialized AI assistants you can create with specific instructions, knowledge, and capabilities - similar to Custom GPTs.
What Are Custom Agents?
Custom Agents are AI assistants configured for specific tasks:
- Specialized instructions: Define how the agent behaves
- Custom knowledge: Attach files for reference
- Consistent behavior: Same instructions every time
- Shareable: Share agents with team or community
Creating a Custom Agent
Basic Setup
-
Open Agent Creation
- Go to Chat Tools
- Click "Create Agent"
- Choose to create new agent
-
Describe Your Agent
- Write what you want the agent to do
- Be specific about its purpose
- Describe capabilities and focus areas
-
AI Enhancement (Optional)
- Let AI enhance your description
- AI generates optimized instructions
- Or manually continue without enhancement
-
Configure Agent
- Name: Give your agent a clear name
- Description: Detailed instructions
- Icon: Choose emoji or upload thumbnail
- Files: Attach reference documents (optional)
-
Save and Use
- Save your agent
- Start using it in chats
- Edit anytime
Agent Configuration
Name
- Clear, descriptive name
- Indicates agent's purpose
- Examples: "Marketing Copy Writer", "Code Reviewer", "Research Assistant"
Description
The description contains your agent's instructions:
- What the agent does
- How it should respond
- Specific guidelines
- Tone and style
- Knowledge areas
Icon
- Choose an emoji that represents the agent
- Or upload a custom thumbnail image
- Helps identify agent quickly
File Attachments
Attach files to give your agent knowledge:
- Documents: PDF, DOC, DOCX, TXT
- Data: CSV, JSON
- Code: Any text-based code files
- Other: SRT, and more
The agent can reference these files when responding.
Using Custom Agents
Start a Chat with Agent
- Select your custom agent
- Start new chat session
- Agent follows its instructions automatically
Agent Behavior
- Follows instructions you provided
- References attached files
- Maintains consistent personality
- Uses full conversation context
Switch Between Agents
- Select different agent in same project
- Start new session with different agent
- Each agent maintains its own context
Agent Library
Your Agents
- View all your created agents
- Edit agent configurations
- Delete agents
- Organize agents
Shared Agents
- Access agents shared with you
- Use team/organization agents
- Import agents from community
Sharing Agents
Share with Team
- Open agent settings
- Click "Share"
- Set permissions
- Share link with team members
Publish to Library
- Share agents with broader community
- Others can import and use
- Build reputation as agent creator
Import Agents
- Import agents shared by others
- Use community-created agents
- Customize imported agents
Agent Use Cases
Content Creation
Marketing Writer Agent
- Brand voice and guidelines
- Product information
- Target audience details
- Writing style preferences
Code Development
Code Review Agent
- Coding standards
- Best practices
- Security guidelines
- Language-specific rules
Research
Research Assistant Agent
- Research methodology
- Citation format
- Analytical approach
- Subject area expertise
Business
Business Analyst Agent
- Company information
- Industry knowledge
- Analysis frameworks
- Reporting format
Customer Support
Support Agent
- Product knowledge
- Company policies
- Response templates
- Troubleshooting guides
Best Practices
Creating Effective Agents
Clear Instructions:
- Be specific about what agent should do
- Include examples of desired behavior
- Define boundaries and limitations
- Specify tone and style
Useful Files:
- Attach relevant reference materials
- Keep files focused and relevant
- Update files as information changes
- Don't overload with too many files
Testing:
- Test agent with various prompts
- Refine instructions based on results
- Iterate until behavior is consistent
- Get feedback from users
Managing Agents
Organization:
- Use clear naming conventions
- Create agents for specific purposes
- Don't make agents too general
- Delete unused agents
Maintenance:
- Update instructions as needs change
- Refresh attached files periodically
- Monitor agent performance
- Collect user feedback
Limitations
- Maximum 100 memories per user (applies to memory system, not agents)
- File attachment size limits apply
- Agent follows general AI model limitations
- Context window limits apply to conversation length
Tips for Success
- Start Simple: Create basic agent first, refine later
- Be Specific: Clear instructions = better results
- Test Thoroughly: Try various scenarios
- Iterate: Improve based on actual usage
- Share Knowledge: Share successful agents with team
- Update Regularly: Keep instructions and files current