Sibling Agents
The Agents tab on an agent's configuration page lets you link your agent to other agents in your workspace, creating a team of specialists that can hand conversations off to each other automatically.
Accessing the Agents Tab
- Go to Agents in the sidebar
- Click the agent you want to configure
- Click the Agents tab (bottom of the left menu)
What are Sibling Agents?
A sibling agent is another agent in your workspace that the current agent knows about and can route conversations to. When agents are linked as siblings, Orki's coordinator watches every conversation and automatically transfers the customer to whichever sibling is best suited for the next step — no trigger keywords, no manual setup per handover reason.
Think of it like a team of specialists at a reception desk: the customer talks to one person, and whenever the topic changes, the receptionist quietly waves over a colleague who's a better fit. The customer experience stays continuous — they don't have to repeat themselves, and they don't see a "transfer" happening.
Why Use Sibling Agents?
Large, monolithic agents with huge system prompts and dozens of tools tend to get confused. Breaking your workflows into smaller, focused agents and linking them as siblings gives you:
| Benefit | Explanation |
|---|---|
| Focused prompts | Each agent has a short, clear system prompt for one domain |
| Smaller tool sets | Each agent only sees the tools relevant to its job, reducing mistakes |
| Easier iteration | Tune one agent's personality or knowledge without breaking others |
| Specialized knowledge bases | Attach different knowledge bases to different agents |
| Scales to complex orgs | Sales, support, billing, and onboarding can each have their own agent |
Example: A Three-Agent Team
Imagine an online fashion store:
- Concierge Agent — greets customers, answers general questions, and routes them
- Product Advisor Agent — knows the catalog, helps with sizing, styling, and product search
- Order Support Agent — handles tracking, returns, refunds, and delivery issues
You would link all three as siblings of each other. A conversation might flow like this:
- Customer: "Hi, I'm looking for a dress for a wedding" → Concierge greets and passes to Product Advisor
- Customer: "Love the blue one! By the way, my last order is stuck" → Coordinator moves the conversation to Order Support
- Customer: "Thanks, all sorted. Can you also show me matching shoes?" → Coordinator moves the conversation back to Product Advisor
All of this happens automatically. The customer sees a single, seamless conversation.
Linking Sibling Agents
- Open the agent you want to configure
- Click the Agents tab
- You'll see two sections:
- Assigned — agents currently linked as siblings
- Available — other agents in your workspace that can be linked
- Flip the switch next to any agent to add or remove it as a sibling
- Use the search box to filter when you have many agents
Sibling links are per agent. If you want two agents to know about each other, you must add the link on both sides. For example, if you link Product Advisor as a sibling of Concierge, you should also open Product Advisor and add Concierge as one of its siblings.
For a small team of 2–4 agents, it's common to link every agent to every other agent, so the coordinator always has full flexibility.
Sibling Agents vs. Handover
Orki now has two distinct ways conversations leave an agent:
| Feature | Purpose | Goes To |
|---|---|---|
| Sibling Agents | Automatic, continuous routing between AI specialists | Another AI agent |
| Handover | Escalation when the AI cannot or should not continue | A human team |
Handover is still the right choice when the customer is frustrated, explicitly asks for a human, or the conversation needs a human touch. Sibling routing is for moving between AI specialists without ever leaving automation.
Earlier versions of Orki let you assign a handover trigger directly to another AI agent. That option has been removed — use Sibling Agents for AI-to-AI routing, and keep handover triggers focused on escalating to human teams.
Best Practices
Keep Each Agent Focused
Split by domain (sales, support, billing, onboarding) or by workflow (discovery, purchase, post-purchase). Avoid creating agents that overlap.
Mirror Your Links
When agent A should be able to transfer to agent B, make sure B can also transfer back to A. The coordinator can only route along links you've explicitly defined.
Start Small
Begin with 2–3 agents. Add more only when you see specific conversations that would benefit from a separate specialist.
Give Each Agent a Clear Description
The coordinator uses each agent's description to decide when to route to it. A description like "Handles order tracking, returns, and delivery issues" is much more useful than "Support agent".
Test in the Playground
Use the Playground to run scenarios that should trigger a sibling transfer, and confirm the coordinator moves the conversation to the right agent.
Next Steps
- Handover Settings - Escalate to human teams when needed
- Personality - Configure each agent's voice and instructions
- Knowledge Base - Give each agent its own specialized KB
- Tools - Give each agent a focused set of tools