Multiple Agents

StrikeUp routes each user message to the most appropriate specialist agent automatically — no manual routing logic required.

Automatic Agent Selection

StrikeUp automatically selects the most appropriate agent for each user message. You don’t write any routing rules — just enable the agents you need and configure them for your use case.

1

User sends message

“How do I request a refund?”

2

Right agent is selected

Support agent handles it

3

Agent responds

Uses refund knowledge + APIs

The Four Agents

Support Agent

Product support and issue resolution

Resolves user issues, explains product features, and guides users through configurations using your connected APIs and knowledge base. Ideal for customer-facing support workflows.

Best for

• ✓ Answering questions about your product

• ✓ Troubleshooting user-reported issues

• ✓ Guiding users through processes step by step

• ✓ Escalating complex issues using human handoff

How to configure

• → Set tone of voice (formal, friendly, empathetic)

• → Add branding instructions (company name, product names)

• → Define behavior rules (what to do / avoid)

• → Assign relevant skills (e.g. Refund Handler, Onboarding Guide)

Copilot Agent

Action executor and reasoning assistant

The primary agent for taking real actions through your APIs. Executes tasks, writes content, and reasons through complex requests. Best for internal tools and power-user workflows.

Best for

• ✓ Executing multi-step API workflows on behalf of users

• ✓ Writing, editing, or summarizing content

• ✓ Reasoning through business logic or configurations

• ✓ Combining multiple API calls into a single response

How to configure

• → Define which tools to prioritize for your use case

• → Set confirmation behavior for destructive actions

• → Configure tone for internal vs customer-facing use

• → Assign skills to extend behavior for specific domains

Analytics Agent

Data retrieval and business intelligence

Specializes in fetching and interpreting business data. Returns structured charts and tables when appropriate. Best for dashboards, reporting, and data-driven decision support.

Best for

• ✓ Querying business metrics and KPIs

• ✓ Generating charts and data visualizations

• ✓ Comparing performance across time periods or segments

• ✓ Surfacing insights from raw API data

How to configure

• → Connect your reporting or analytics APIs

• → Define preferred chart types and data formats

• → Set context about your metrics (what each field means)

• → Add skills for domain-specific calculations

General Agent

Fallback for cross-domain requests

Handles requests that span multiple domains or don’t fit a specialized category. Acts as a flexible fallback for open-ended, exploratory, or mixed-intent conversations.

Best for

• ✓ Mixed-intent queries covering multiple topics

• ✓ Exploratory or open-ended questions

• ✓ Fallback when no specialized agent matches

• ✓ Simple conversational exchanges

How to configure

• → Keep instructions broad and flexible

• → Avoid over-specialization for this agent

• → Use skills to add lightweight domain knowledge

• → Set a neutral, professional tone

Setting Up Agents

1

Open Widget Settings

Navigate to Widgets → select your widget → Agents tab.

2

Enable the Agents You Need

Each agent can be independently enabled or disabled. Enable Support and Copilot for most use cases. Add Analytics if your widget connects to reporting APIs.

3

Configure Each Agent

Set the Tone of Voice, Behavior Instructions (what the agent should always / never do), and Branding Instructions (your company name, product terminology).

4

Assign Skills (Optional)

From the agent card, click Manage Skills to assign pre-built skills. Skills extend the agent’s behavior for specific situations without bloating its base instructions.

5

Test Routing

Open the widget preview and try different types of questions. Check AI Logs to see which agent handled each request.

Agent Policies

Each agent has independent policies you can configure per widget:

Pro tip

Start with Support + Copilot enabled and General as fallback. Add Analytics only when your widget is connected to data APIs. Only enable agents that are relevant to your use case — focused configurations deliver better, more consistent results.