> ## Documentation Index > Fetch the complete documentation index at: https://docs.banya.ai/llms.txt > Use this file to discover all available pages before exploring further. # Settings > Explains each tab of the CodePilot IDE settings screen so that even first-time users can understand it. Click the **gear (⚙) icon** at the top right of the chat panel to open the settings screen. Settings screen When connected to an Organization, items the administrator has deployed as **Required** cannot be changed. These are the items marked with a lock icon. *** ## Account Account settings This screen shows your login status and the Organization you belong to, and handles joining an Organization and selecting a Project. | Item | Description | | -------------------------- | ----------------------------------------------------------------------- | | Profile | Login account, Organization membership, authentication status | | Organization connection | Register an Organization API key (`cpk_...`) and leave the Organization | | Project | Select the Project to work on | | Sync organization settings | Re-fetch the latest policies from the server | For detailed steps, see the [Organization connection](/en/getting-started/organization) doc. *** ## General General settings ### Language / Theme / Default Mode | Item | Description | | ------------ | --------------------------------------- | | Language | AI response language (Korean / English) | | Theme | Dark / Light / Follow VS Code setting | | Default Mode | The mode selected when a chat starts | ### Auto-execution permissions Determines whether the AI **asks for confirmation or runs immediately** before it modifies files or executes commands. | Setting | When ON | When OFF | | ----------------- | ---------------------------------- | -------------------------------- | | Auto-run tools | The AI runs tools immediately | Asks for approval before running | | Auto-update files | Applies changes immediately | Approve after reviewing the diff | | Auto-run commands | Runs terminal commands immediately | Confirm before running | If you're using it for the first time, start with **Auto-run tools and Auto-run commands turned OFF**. We recommend turning them on only after you've built trust by observing what the AI does. ### Other * **Inline code completion** — Tab completion while typing (OFF by default) * **Next-task suggestions** — Suggests follow-up tasks after a task completes * **Web search** — Reflects web search results in answers * **Credential protection** — Inspects responses so that secret keys aren't exposed * **Export / import settings** — Back up and restore settings as a JSON file *** ## AI Models AI Models settings ### Selecting the main model Choose the default model to use for chat. | Type | Characteristics | | ----------- | ------------------------------------------------------ | | Cloud model | High performance, requires internet, incurs cost | | Local model | Free, works offline, speed varies with your PC's specs | When connected to an Organization, only models approved by the administrator appear in the list. ### Response options | Option | Description | | ---------- | ---------------------------------------------- | | Streaming | Displays the answer as it's generated | | Thinking | Reason more deeply (Low / Medium / High / Max) | | Multimodal | Supports image and PDF attachments | ### Registering a personal API key Besides the Organization's models, you can register a personal API key and use it at your own expense. Enter the key issued by the provider on this screen. ### Assigning models by purpose You can assign a different model to each type of task. Assigning a lightweight model to simple tasks improves both cost and response speed. | Purpose | Recommended | | ----------------------- | ------------------------------------- | | Conversation compaction | Lightweight model | | Command execution | Main model | | Intent analysis | Lightweight model | | Error recovery | A model different from the main model | | Inline code completion | Lightweight local model | *** ## MCP Servers MCP Servers settings Connect external tools such as GitHub, Jira, and in-house systems so the AI can operate them directly. * **Local connection** — Tools that run on your own PC * **Remote connection** — An external MCP server URL Authentication supports API keys or OAuth. If your Organization administrator pre-registers MCP servers, you can use them right away with no additional setup. *** ## RAG (Internal Documents) RAG settings View the list of internal documents the administrator registered in the CodePilot Admin console. During chat, relevant documents are automatically searched and reflected in answers. Since document registration and indexing are performed by the administrator, in the IDE you only view which documents are connected. *** ## Build/Test Build/Test settings Specify the verification commands to run automatically after writing or fixing code. | Project | Example commands | | ------------- | ------------------------------------------- | | Node.js | `npm test`, `npm run lint`, `npm run build` | | Python | `pytest` | | Java (Maven) | `mvn verify` | | Java (Gradle) | `./gradlew test` | If left empty, the project type is detected and commands are chosen automatically. If verification fails, the AI analyzes the cause, applies a fix, and verifies again. *** ## Hot Load Hot Load settings Register frequently used tasks as keywords, and when you type a sentence containing that keyword in the chat, the registered command runs. ``` Keyword: deploy, release Description: Run a production deployment Command: ./scripts/deploy.sh production ``` You can also specify a completion condition (exit code, output string, or file existence), the number of retries, and the behavior on failure. *** ## Hooks Insert shell commands at specific points in the workflow. Use them for audit logging, integration with in-house tools, and more. Hooks settings | Event | When it fires | | --------------------- | --------------------------------------- | | Before tool execution | Just before the AI runs a tool | | After tool execution | After a tool finishes running | | On tool failure | When a tool execution fails | | Before LLM call | Just before a request is sent to the AI | | On session start | When a new conversation starts | *** ## Agents Register subagents that take on specific roles. Once registered, you can invoke them in chat with **`@agent-name`**. Agents settings | Item | Description | | ------------- | ------------------------------------------- | | Name | The call name that follows `@` | | Description | What this agent does | | System prompt | The agent's role and instructions | | Allowed tools | The tools this agent can use | | Read-only | Blocks file modification and deletion tools | Your Organization administrator can also deploy company-wide shared agents. *** ## Skills (Skills and Rules) Skills settings Register, in Markdown, the team rules the AI must follow when writing code. ``` Write all React components as function components, and define Props types using an interface. ``` The categories are divided into Global Rules, Version Control, Coding Style, Project Architecture, Dependency Policy, and Database. | Type | Description | | ------------------ | --------------------------------------------------------------- | | Local rules | Applied only on your own PC | | Organization rules | Registered by the administrator and applied to all team members | **Required** items among the Organization rules cannot be modified or turned off by individuals. *** ## Security Security settings Specify the commands the AI must not run and the files it must not access. ### Blocked commands Blocks commands that destroy the system or are irreversible. ``` rm -rf / DROP TABLE git push --force chmod 777 ``` ### Protected files / exclude patterns Specify the paths the AI is blocked from reading or modifying. ``` .env ← API keys, passwords .env.local secrets/ ← certificates, secret keys config/secrets.yml ``` Paths listed in `.gitignore` are automatically skipped during file exploration. To block them completely, however, add them directly to the exclude patterns. Security settings are the last line of defense that prevents the AI from accidentally touching important files. Be sure to add your `.env` file to the exclude patterns. *** ## Usage View the memory used, the number of AI calls, the number of tool executions, and the number of files referenced in the current session. *** ## Per-Project Settings When you select a Project, the **Team default settings** are also shown, read-only, on the settings screen. For any item the Project doesn't set separately, the Team default settings apply as-is. Project security settings The Team default settings apply. The Organization-wide settings registered by the administrator are used as-is. This means the administrator hasn't created any Projects yet, or you haven't been added as a member of the Project. Contact your administrator. No. Your conversation is preserved, and the new Project's settings apply from the next message onward. Organization rules take precedence. Use local rules to complement the Organization rules.