How It Works
1
Task Identification
You’re chatting with an Expert. While working on your request, the Expert identifies a subtask that requires a different domain of expertise — for example, your full-stack Expert needs a data analysis Expert to handle a statistical task.
2
Automatic Delegation
The Expert automatically selects the right specialist and initiates the delegation. You’ll see the process in the conversation, including which Expert was chosen and the specific task being delegated.
3
Subtask Execution
The delegated Expert takes over the task in the background, using its own Skills and tools to complete the work.
4
Result Return
Once the subtask is complete, the results are automatically returned to the primary Expert. The primary Expert integrates the results into the current conversation and continues advancing your task.
Typical Scenarios
Cross-Domain Collaboration
Your coding Expert needs data visualization during development → delegates to a chart design Expert
Language Translation
Your Chinese writing Expert needs to translate a technical document → delegates to a translation Expert
Professional Review
Your frontend Expert has completed page development → delegates to a security Expert for a code security audit
Content Creation
Your technical Expert has written feature documentation → delegates to a marketing Expert to polish it into user-facing copy
Delegation Transparency
The entire delegation process is fully transparent to you. You can clearly see in the conversation:- Delegation initiated: Which Expert the primary Expert delegated to, along with the specific task description
- Execution process: The delegated Expert’s work process (shown as a summary)
- Results returned: The subtask completion results and how the primary Expert integrates them
Delegation Hierarchy
To ensure conversation security and controllability, delegation follows a layered tool permission mechanism:- Primary Expert (L1): Has full tool permissions, including core capabilities like plan creation and delegation scheduling
- Delegated Expert (L2): Has tool permissions needed for the subtask, but certain core management capabilities are reserved for the primary Expert
Conversation content and tool calls generated by delegation also consume your credits. Exact consumption is shown in real time within the product.
Delegatable Experts
The range of Experts available for delegation depends on:- Experts you’ve used: Experts you’ve previously chatted with can be delegated to
- Publicly listed Experts: Published Experts on the marketplace can also be delegated to
Delegation Authorization Settings
You can manage delegation behavior in your personalization settings:- Allow delegation: A global toggle controlling whether Experts can initiate delegation
- Target allowlist: Specify which Experts can be delegated to
FAQ
Does delegation affect the main conversation's context?
Does delegation affect the main conversation's context?
No. Delegation is executed in an isolated context. The delegated Expert only sees the subtask description passed by the primary Expert — it cannot access your full conversation history. Once the subtask is complete, results are returned to the main conversation as a summary.
Can I prevent Expert from auto-delegating?
Can I prevent Expert from auto-delegating?
Yes. You can disable delegation in your personalization settings. When disabled, the Expert won’t initiate any delegation and will handle all tasks itself.
What happens if a delegated subtask fails?
What happens if a delegated subtask fails?
The primary Expert receives failure feedback and attempts an alternative approach — it may handle the subtask itself, or inform you of the issue and ask for your guidance.
Does delegation use the delegated Expert's memory?
Does delegation use the delegated Expert's memory?
The delegated Expert can access its own memory and Skills when executing the subtask. This is precisely the value of delegation — leveraging its specialized expertise and accumulated knowledge.

