Skip to main content

Core Workflows

This document explains typical workflows for completing daily development tasks in Tuanjie AI, and maps concepts from common AI editor documentation to help users migrating from other tools get up to speed.

Main Things You Do in Tuanjie AI​

  1. Dialog & Tasks: Describe your requirements in natural language, and the model generates or modifies code, explains projects, and troubleshoots issues. This corresponds to "Chat" or "Ask" capabilities in many products.
  2. Structured Automation: Turn repetitive workflows into reusable entry points through slash commands, custom Commands, and Skills.
  3. Multi-step Delegation: Delegate specialized work to Subagents, which handle investigations, testing, or specialized refactoring in independent contexts, then report back to the main session.

Model Selection (Similar to "Select Model/Select Mode")​

Tuanjie AI balances speed and depth through model tiers like Basic and Core. You can switch before sending messages. See Models for details.

Rule of thumb: Use Core for complex, multi-step implementations and debugging; Basic is suitable for quick Q&A and lightweight modifications.

Correspondence with "Agent Mode"​

Many documents emphasize "Agent Mode": the model can call tools in multiple steps, browse codebases, and continuously execute tasks. In Tuanjie AI, this capability is distributed across:

  • Main Conversation: Continuous multi-turn dialog in Cowork, IDE extensions, or CLI, combined with @ mentions and MCP to pull in external context.
  • Subagents: Explicitly enable or @ specify expert agents, suitable for specialized subtasks that require isolated contexts. See Subagents for details.

Inline Editing and "Cmd K" Style Workflows​

If you're used to selecting code in an editor and initiating "inline editing," in Tuanjie AI IDE integration this is typically done by: Select code → Describe the modification in the Tuanjie AI dialog, with @ file path or screenshots (see Screenshot Analysis). CLI users can use @ to reference files or directories in prompts directly—see Codely CLI · Referencing Files and Directories.