Custom AI Agents

Build specialized AI agents in minutes without coding. Describe what you need, and your custom agent is ready instantly.

What is a Custom Agent?

A Custom Agent is a specialized AI powered by Neuro that you design for a specific purpose. Instead of asking a general-purpose AI, you have agents trained on your exact requirements.

Think of Custom Agents as:

• Specialized expertise — Focused on one job
• Instant creation — No coding required
• Reusable — Use anytime via chat
• Shareable — Build once, share with others
• Community-driven — Discover amazing agents built by others

---

Quick Start: Create Your First Agent

Step 1: Access Custom Agent Creation

Option 1: From Homepage
Click "Create Custom Agent" button

Option 2: From Sidebar
Click "+ New" → "Custom Agent"

Option 3: In Chat
Type @ → "Create New Agent"

Step 2: Define Your Agent

Simple Format:

Title: [Agent Name]
Role: [What is this agent?]
Purpose: [What does it do?]
Output: [How should it respond?]

Example:

Title: Weekly AI News Analyst
Role: AI industry analyst
Purpose: Generate weekly summaries of funding trends, 
        product launches, and policy changes
Output: Organized bullet points with sources

Step 3: Test & Refine

1. Click "Preview" to test
2. Ask sample questions
3. Review responses
4. Adjust instructions if needed
5. Click "Update" to save

Step 4: Start Using

Use in three ways:
- Direct chat with agent
- @mention in conversations
- Invite others to use it

---

Creating Effective Custom Agents

The Three Key Components

Every good Custom Agent needs:

1. Role Definition

Clear professional identity

Good:
"You are an expert financial analyst specializing 
in tech startup valuations. You have 15 years of 
experience in venture capital."

Too vague:
"You are helpful and knowledgeable"

Why it matters:
- Tells agent its expertise area
- Sets appropriate tone
- Defines scope of knowledge

2. Core Functions

What the agent actually does

Good:
"Your job is to:
1. Analyze financial statements
2. Benchmark against industry standards
3. Identify growth opportunities
4. Provide valuation estimates
5. Flag risk factors"

Why it matters:
- Keeps agent focused
- Prevents scope creep
- Ensures consistency

3. Output Standards

How the agent presents information

Good:
"Format responses as:
- Executive Summary (1 paragraph)
- Key Findings (5 bullet points)
- Data Table (if numeric)
- Recommendations (3 items)
- Sources (citations required)"

Why it matters:
- Users know what to expect
- Consistent quality
- Professional presentation

---

Agent Examples by Type

Example 1: Research Agent

Title: Market Research Analyst

Role: 
Market research specialist focused on B2B SaaS 
trends and competitive intelligence

Purpose:
Generate comprehensive market analysis reports 
including market size, growth trends, key players, 
and emerging opportunities

Output Format:
1. Market Overview (size, growth rate, segments)
2. Competitive Landscape (top 5 players)
3. Emerging Trends (3-5 key trends)
4. Opportunities (white space analysis)
5. Risk Factors (5 potential challenges)

Conversation Starters:
- "Analyze the AI automation market"
- "Who are the top players in no-code automation?"
- "What's the market opportunity in enterprise AI?"

Example 2: Writing Agent

Title: LinkedIn Content Creator

Role:
Professional content writer specializing in 
LinkedIn posts for tech executives

Purpose:
Create engaging LinkedIn content that drives 
engagement and establishes thought leadership

Output Format:
1. Hook (compelling first line)
2. Story or insight (3-4 sentences)
3. Key point (1 main takeaway)
4. CTA (call-to-action)
5. Hashtags (3-5 relevant)

Tone: Professional but conversational
Length: 150-250 words
Include: Data points or personal insights

Conversation Starters:
- "Write about AI trends in enterprise software"
- "Create a post about remote work benefits"
- "Generate thought leadership content on my expertise"

Example 3: Analysis Agent

Title: Investment Due Diligence Analyst

Role:
Experienced investment analyst performing 
startup due diligence

Purpose:
Evaluate startup investment opportunities by 
analyzing financial projections, market fit, 
team quality, and risk factors

Output Format:
1. Investment Recommendation (pass/caution/strong interest)
2. Financial Analysis (3 key metrics)
3. Market Assessment (size, timing, competition)
4. Team Evaluation (founders, experience gaps)
5. Risk Assessment (top 3 risks)
6. Questions for Founders (5-7 due diligence questions)

Analysis Framework: Use venture capital standards

Conversation Starters:
- "Should we invest in this company?"
- "Analyze this startup's financials"
- "Evaluate market fit for this business"

Example 4: Productivity Agent

Title: Project Planning Assistant

Role:
Project management expert specializing in 
agile methodologies

Purpose:
Help plan and structure projects using best 
practices in project management

Output Format:
1. Project Overview (scope, timeline, budget)
2. Milestone Timeline (key deliverables and dates)
3. Risk Register (risks and mitigation)
4. Resource Plan (team, skills, allocation)
5. Success Criteria (how success is measured)

Structured Output: Use tables and clear sections

Conversation Starters:
- "Create a project plan for this initiative"
- "What's a realistic timeline for this?"
- "Help me identify project risks"

Example 5: Learning Agent

Title: Machine Learning Tutor

Role:
Patient ML educator specializing in helping 
beginners understand complex concepts

Purpose:
Teach machine learning concepts through clear 
explanations, real-world examples, and hands-on 
projects

Output Format:
1. Simple Explanation (no jargon)
2. Real-World Example (relatable use case)
3. How It Works (step-by-step)
4. Practical Project (something to build)
5. Next Steps (what to learn next)

Teaching Style: 
- Assume no prior knowledge
- Use analogies from everyday life
- Include Python code examples
- Provide practice problems

Conversation Starters:
- "Explain neural networks simply"
- "How do I get started with machine learning?"
- "Create a beginner ML project for me"

---

Managing Custom Agents

View Your Agents

Location: Custom Agent Store → "My Agents"

Shows:
- Agents you created
- Your favorite agents
- Favorite count
- Last modified date
- Quick actions

Edit an Agent

Steps:
1. Find agent in "My Agents"
2. Click three-dot menu
3. Select "Edit"
4. Modify instructions, name, or settings
5. Test in preview
6. Click "Save Changes"

You can edit:
- Title and description
- Role definition
- Functions and purpose
- Output format
- Conversation starters
- Privacy settings

Delete an Agent

Steps:
1. Find agent in "My Agents"
2. Click three-dot menu
3. Select "Delete"
4. Confirm deletion

⚠️ WARNING: Deletion is permanent and cannot be undone

Best Practice: Export conversation history before 
deleting to preserve any useful data

Favorite an Agent

When viewing Custom Agent Store:
1. Click heart icon on agent card
2. Agent appears in "My Agents"
3. Can @mention in conversations
4. Quick access for repeated use

---

Using Custom Agents

Three Ways to Use

1. Direct Chat

Click on agent → Start conversation

Use when:
- Need focused conversation with one agent
- Testing agent behavior
- Deep work with specific expertise
- Detailed back-and-forth discussion

2. @Mention in Conversation (soon)

In chat, type: @AgentName

Use when:
- Need multiple agents in one conversation
- Building on previous context
- Quick question for specific agent
- Combining agent outputs

Example:
"@MarketResearch analyze this market
@CompetitiveIntelligence compare to competitors
Now @SynthesisAgent combine these into a report"

3. Collaboration (soon)

Invite others to use your agent:
1. Click "Share"
2. Choose sharing option
3. Send link or publish to store

Collaborate when:
- Team needs consistent analysis
- Want community feedback
- Building on shared expertise
- Teaching or mentoring

---

Sharing Custom Agents

Sharing Options

Private (Only Me)

Who can see: Only you
Who can use: Only you
Use when: Personal agents, work-in-progress

Best for:
- Testing new agents
- Personal productivity tools
- Private analysis for your work

Link Sharing

Who can see: Anyone with the link
Who can use: Anyone with the link
Privacy: Your conversation history is private
          Others can't see your chats

Use when: Sharing with specific people
         Testing with feedback group
         Small team collaboration

Share method:
- Generate shareable link
- Send via email, Slack, etc.
- Revoke access anytime

Public (Custom Agent Store)

Who can see: Everyone
Who can use: Everyone
Discovery: Appears in trending/categories
Rating: Users can favorite and review

Use when: Want to share with community
         Building reputation
         Creating useful tools
         Contributing to ecosystem

Publishing steps:
1. Edit agent
2. Click "Share"
3. Select "Public - Custom Agent Store"
4. Add description and tags
5. Publish

What Gets Shared

When you share, others see:
✓ Agent name
✓ Description
✓ Category
✓ Conversation starters
✓ Author name

They CANNOT see:
✗ Detailed instructions
✗ Your conversation history
✗ Your chats with agent
✗ Sensitive configuration

---

Best Practices for Custom Agents

✅ Do’s

✓ Be specific with role definition
  "Expert financial analyst with 15 years in VC"
  Not: "Helpful AI"

✓ Define clear output format
  Include structure, style, length expectations

✓ Test thoroughly
  Try various questions before publishing

✓ Include conversation starters
  Guide users on how to use agent effectively

✓ Set realistic boundaries
  Define what agent will and won't do

✓ Use concrete examples in instructions
  "Format as: Executive Summary + Key Points + Data"

✓ Update based on feedback
  Refine agent over time

✓ Add clear description
  Help others understand purpose

❌ Don’ts

✗ Don't create vague agents
  "General purpose AI" is too broad

✗ Don't overload with functions
  Focus on core purpose

✗ Don't share sensitive information
  In instructions or examples

✗ Don't publish untested agents
  Test thoroughly first

✗ Don't abandon shared agents
  Update if you get feedback

✗ Don't create duplicate agents
  Check if it already exists

✗ Don't mislead about capabilities
  Be honest about what it does

✗ Don't ignore user feedback
  Learn from how people use it

---

Optimizing Agent Instructions

Conversation-Based Optimization

In Create dialog:
1. Describe what you want naturally
2. System suggests improvements
3. Refine iteratively
4. Test in preview

Example:
You: "I need an agent that analyzes companies 
and tells me if they're good investments"

System improves:
"I need an agent that evaluates startup 
investment opportunities by analyzing:
- Financial projections and runway
- Market size and competitive positioning
- Founding team background and experience
- Key risks and mitigation strategies
Output structured analysis with recommendation"

Manual Optimization

After testing, directly edit:

1. Role clarity
   Make expertise more specific

2. Function boundaries
   Define what agent will/won't do

3. Output structure
   Improve presentation

4. Tone adjustment
   More formal, casual, technical, etc.

5. Example expansion
   Add more guidance

Test after each change in preview

Conversation Starters

Help users understand how to use agent

Good starters:
- "Analyze this company as an investment"
- "What are key risks in this market?"
- "Create a project plan for this initiative"

Bad starters:
- "Chat with me"
- "Ask me anything"

Function: Guide users to best use cases

---

Advanced Agent Techniques

Agent Chaining

Combine multiple agents for complex tasks

Example: Investment Report
1. @FinancialAnalyst analyzes numbers
2. @MarketResearcher studies market opportunity
3. @CompetitiveIntel compares to competitors
4. @SynthesisAgent combines into report

When to use:
- Complex multi-step analysis
- Combining different expertise
- Producing comprehensive outputs

Agent Specialization

Create focused agents for specific needs

Good approach:
- Investment Analyst
- Market Research Analyst
- Competitive Intelligence Analyst
- Legal Compliance Analyst

Avoid:
- One agent doing everything
- Overlapping responsibilities

Benefit:
- Better quality
- Easier to improve
- Reusable across projects

Agent Versioning

Keep multiple versions for different uses

Example:
- MarketAnalysis_Detailed (comprehensive)
- MarketAnalysis_Brief (executive summary)
- MarketAnalysis_Financial (numbers focused)

Benefit:
- Reuse without modification
- Different contexts
- Different user needs

---

Security & Privacy

Your Data is Safe

✓ Your conversations are private
  Creator cannot see your chats
  
✓ Agent instructions stay private
  Others can use, but not see details
  
✓ Sharing is selective
  You control who sees what
  
✓ Deletion is permanent
  No recovery possible
  
✓ No data leakage
  Each user's history is isolated

Using Other People’s Agents

When you use a shared agent:

Your privacy:
✓ Your conversation history is private
✓ Creator cannot view your chats
✓ Your data is not shared
✓ Complete isolation

Their privacy:
✓ You can't see their instructions
✓ You can't see their conversation data
✓ You get agent functionality only

---

Common Custom Agent Examples

By Department

Sales

• Prospect Research Agent
• Sales Email Generator
• Deal Analysis Agent
• Competitor Intelligence Agent

Marketing

• Content Calendar Planner
• Social Media Post Creator
• Market Research Analyst
• Campaign Strategist

Product

• Feature Requirement Writer
• Roadmap Strategist
• Customer Feedback Analyzer
• Competitive Feature Analyzer

Operations

• Process Documentation Agent
• Project Planning Assistant
• Meeting Summarizer
• Workflow Optimizer

Finance

• Financial Analysis Agent
• Investment Due Diligence Agent
• Budget Planner
• Financial Forecasting Agent

HR

• Job Description Writer
• Interview Question Generator
• Onboarding Specialist
• Performance Review Analyst

---

Testing Your Agent

What to Test

Accuracy:
□ Does it answer correctly?
□ Is information accurate?
□ Are examples relevant?

Consistency:
□ Same type of question = similar responses?
□ Tone consistent?
□ Format consistent?

Quality:
□ Output well-organized?
□ Appropriate depth?
□ Useful recommendations?

Boundaries:
□ Stays within scope?
□ Declines out-of-scope questions appropriately?
□ Handles edge cases well?

Usefulness:
□ Actually helpful?
□ Better than general AI?
□ Worth sharing?

Sample Test Questions

Basic functionality:
- Simple question in core area
- Follow-up question
- Question at domain boundary

Edge cases:
- Ambiguous question
- Request outside scope
- Conflicting requirements

Stress test:
- Complex multi-part question
- Request for unusual format
- Deep dive into specific area

---

FAQ

How do I create my first custom agent?
Click “Create Custom Agent,” describe what you need, test in preview, then save. No coding required.

Can I edit agents I created?
Yes. Click the three-dot menu, select “Edit,” make changes, test, and save.

What if I delete an agent by mistake?
Deletion is permanent and cannot be undone. Export conversation history before deleting.

Can I edit agents from the store?
Only your own agents. Agents you’ve favorited can be used but not edited.

Will others see my conversations?
No. Your conversation history is completely private. Creators cannot view your chats.

Can I share an agent with my team?
Yes. Click “Share,” choose “Link Sharing” or “Team Only,” and share the link.

How do I make my agent better?
Test it, get feedback, edit instructions, test again. Iterate until you’re happy.

Can I use multiple agents together?
Yes. Use changing button to invoke multiple agents in one conversation for complex tasks.

How specific should my role definition be?
Very specific. “Expert financial analyst with 15 years in venture capital” is better than “helpful analyst.”

Can I download agent instructions?
You can export conversations. Instructions remain with the agent.

Should I make my agent public?
Only if it’s well-tested and genuinely useful to others. Start private and improve first.

What makes a good agent name?
Clear and descriptive: “LinkedIn Content Creator” not “Agent1”

---

Getting Started

Create Your First Agent

5-minute setup:

1. Click "Create Custom Agent"
2. Give it a clear name
3. Define the persona (2-3 sentences)
4. List core functions (3-5 items)
5. Define output format
6. Test in preview
7. Publish

Then:
- Use directly
- Share with others
- Refine based on feedback
- Build more agents

Build Your Agent Library

Month 1: Create 2-3 core agents
Month 2: Refine and optimize
Month 3: Add specialized agents
Month 4: Share with team
Month 6: Build agent ecosystem

Each agent should solve a specific problem

---

Next Steps

1. Create Your First Agent — Choose a specific purpose
2. Test Thoroughly — Try various questions
3. Refine Instructions — Optimize based on results
4. Use Regularly — Integrate into workflows
5. Share When Ready — Help your team or community

---

Last updated: 2026
Ready to build custom agents? Create your first one today!