Agent mode features autonomous decision-making, codebase awareness, and tool use capabilities. It also offers tools like project search, file editing, and terminal execution to complete end-to-end coding tasks based on developer requirements. Additionally, it supports MCP tool configuration, expanding the AI coding assistant's capabilities and better aligning with developer workflows.
To use Agent mode, update Lingma to version 2.5.0 or above in VS Code or JetBrains IDEs. This feature is still in the preview stage, and we welcome your feedback on it.
The Agent mode offers the following core features:
Project-level changes: Autonomously breaks down tasks and modifies code files in the project based on task description. It also iterates on your project and makes step-by-step changes through multi-turn chats or reverting snapshots.
Automatic codebase awareness: Automatically detects project frameworks, technology stacks, required code files, and error messages from task descriptions. This eliminates the need to add context manually, making task descriptions easier.
Tool use: Autonomously uses more than ten built-in tools, such as file read/write, code queries, and error troubleshooting. It also supports MCP tool configuration and integrates with the ModelStudio MCP Square, allowing developers to configure freely.
Command execution: Autonomously decides which commands to execute. Then it generates these commands and runs them in the terminal, greatly improving efficiency.
Use programming tools
Lingma provides more than ten programming tools for the agent to use autonomously. These tools include file search, file reading, directory reading, semantic symbol retrieval within the project, file modification, getting error messages, and running terminal commands.
The agent decides which tool to use and how to use it without asking the developer. It also plans its next steps based on the returned results.
Execute commands
Agent mode autonomously chooses and runs terminal commands based on current needs. By default, the developer must confirm each command before it runs, as follows:
Click the run button, and the agent will send the command to the IDE Terminal window for execution.
For tasks that run in the background, a "background running" mark will show. The agent will keep working on others tasks and will check or get the results when needed.
If you encounter any issues when running terminal command, see Terminal execution exceptions for solutions.
You can configure an allowed list for automatic execution in the plugin settings. Configured commands can be executed automatically without confirmation. To add multiple commands, separate them with commas.
Call MCP tools
When MCP tools are already configured, the agent autonomously decides whether to call MCP to complete tasks. Before executing each task, it asks you for confirmation. To confirm, click the execute button. The agent will then call the MCP tool and use the returned results as context for the next steps.
For more details on using MCP in Lingma, see MCP.
