Qwen Code CLI reference

Analyze current directory and create initial context file.

Generate project summary based on conversation history.

optional [path]; default .qwen/PROJECT_SUMMARY.md

Replace chat history with summary to save tokens (alias /summarize).

Fast compression without AI; strips old tool outputs and thinking parts.

Resume a previous conversation session.

Generate a one-line session recap now.

Revert project files to the checkpoint before a tool call ran.

Delete a previous session.

Fork the current conversation into a new session.

Spawn a background agent that inherits the full conversation.

Rewind conversation to a previous turn.

Export session history to file.

html, md, json, jsonl

Rename or tag the current session.

Create a git commit with an AI-drafted message.

Clear conversation history and free up context.

Show context window usage breakdown; /context detail shows per-item usage.

Control history display preferences and visibility.

Open an interactive diff viewer showing uncommitted changes and per-turn diffs.

Open a commit history viewer for the workspace (Web Shell only).

Change Qwen Code visual theme.

Turn input area Vim editing mode on/off.

Toggle voice dictation input.

hold, tap, off, status

Manage multi-directory support workspace (alias /dir).

Move this session to a new working directory.

Open dialog to select supported editor.

Open interactive status line preset dialog; with text, generate a command-mode status line via agent.

Configure terminal keybindings for multiline input.

View or change language settings; ui sets UI language, output sets LLM output language.

ui [language], output [language]

List configured MCP servers and tools.

Import MCP servers from Claude configs.

Display currently available tool list.

Open the Skills panel to browse, search, toggle, and launch skills.

Create a reusable project skill from a file, directory, URL, video, or text.

Inspect, pin, archive, or restore inactive project auto-skills.

Switch to plan mode or exit plan mode.

Change the tool-approval mode (current session only).

plan, default, auto-edit, auto, yolo

Review held peer messages; manage trusted controllers.

Switch model used in current session.

Set a lighter model for prompt suggestions.

Set the model used for voice transcription.

Set the vision-bridge model used to transcribe images for a text-only main model.

Set the model used for chat compression.

Set an image-generation-capable model for the built-in image generation tool.

Set reasoning effort for thinking-capable models (mapped and clamped per provider).

low, medium, high, xhigh, max

Choose the output style that shapes how responses are written.

Manage extensions (list, manage, explore, install); changes hot-reload.

list, manage, explore, install

Open the Memory Manager dialog.

Save a durable memory.

Remove matching entries from auto-memory.

Manually run auto-memory consolidation.

Manage Qwen Code hooks.

Reload extension changes (commands, skills, agents, workflows, hooks, MCP/LSP servers) from disk.

Manage permission rules.

Manage subagents.

Manage Arena sessions.

start, stop, status, select (alias choose)

List background tasks.

Inspect workflow runs; cooperatively pause/resume a background run.

Show LSP server status (registered only with --experimental-lsp).

Manage folder trust settings.

Multi-agent code review (12 parallel agents at high effort).

--comment, --fix, --resume, --effort low|medium|high

Coordinate read-only workers and one optional worktree writer.

Run a prompt on a recurring schedule.

Turn a fuzzy intention into a verifiable /goal objective.

Review recent changes and apply safe cleanup edits directly.

Answer questions about Qwen Code usage and configuration.

Ask a quick side question without interrupting the main conversation.

Get a second-opinion review of the conversation above, optionally with a focus.

optional <focus>

Display help information for available commands.

Display version information; /status paths shows session file and log paths.

Open the interactive usage statistics dashboard (Session, Activity, and Efficiency tabs).

model, tools, skills, daily, monthly, export

Show per-model token breakdown and estimated cost.

Export usage statistics to CSV or JSON.

Open settings editor.

Get or set any setting by dot-path key (writes to user settings).

Change authentication method.

Run installation and environment diagnostics.

memory, cpu-profile, rollback

Open full Qwen Code documentation in browser.

Manage IDE integration.

Generate programming insights from chat history.

Set up GitHub Actions.

Submit issue about Qwen Code.

Copy to clipboard: reply (Nth-last), code (by lang), LaTeX, or Mermaid.

Exit Qwen Code immediately.

Ask the agent to plan and submit a Batch API job; it previews the plan before anything is billed.

Update a standalone install from inside a session; honors QWEN_UPDATE_BASE_URL.

Specifies the Qwen model to use for this session.

Alias: -m. Values: Model name

Used to pass a prompt directly to the command. This invokes Qwen Code in a non-interactive mode.

Alias: -p. Values: Your prompt text

Starts an interactive session with the provided prompt as the initial input.

Alias: -i. Values: Your prompt text

Overrides the built-in main session system prompt for this run.

Values: Your prompt text

Appends extra instructions to the main session system prompt for this run.

Values: Your prompt text

Selects the output style that shapes how responses are written for this run.

Values: Style name

Specifies the format of the CLI output for non-interactive mode.

Alias: -o. Values: text, json, stream-json

Specifies the format consumed from standard input.

Values: text, stream-json

Include partial assistant messages when using stream-json output format.

Enables sandbox mode for this session.

Alias: -s

Enables debug mode for this session, providing more verbose output.

Alias: -d

Displays help information about command-line arguments.

Alias: -h

Enables YOLO mode, which automatically approves all tool calls.

Sets the approval mode for tool calls.

Values: plan, default, auto-edit, auto, yolo

A comma-separated list of tool names that will bypass the confirmation dialog.

Values: Tool names

Slash command names to hide/disable (comma-separated or repeated).

Values: Command names

Enables telemetry.

Sets the telemetry target.

Sets the OTLP endpoint for telemetry.

Sets the OTLP protocol for telemetry (grpc or http).

default grpc

Enables logging of prompts for telemetry.

Enables ACP mode (Agent Client Protocol). Useful for IDE/editor integrations like Zed.

Enables the experimental LSP (Language Server Protocol) feature for code intelligence.

Specifies a list of extensions to use for the session.

Alias: -e. Values: Extension names

Lists all available extensions and exits.

Alias: -l

Sets the proxy for the CLI.

Values: Proxy URL

Includes additional directories in the workspace for multi-directory support.

Values: Directory paths

Enables screen reader mode, which adjusts the TUI for better compatibility with screen readers.

Displays the version of the CLI.

Enables logging of OpenAI API calls for debugging and analysis.

Sets a custom directory path for OpenAI API logs.

Values: Directory path

Disable all customizations (context files, hooks, extensions, skills, MCP servers, and more) to isolate problems. Also settable via QWEN_CODE_SAFE_MODE=true.

Resume the most recent session for this project.

Resume a specific session (or choose interactively).

Cap the number of user/model/tool turns in the run; exits with code 53 on overrun.

Settings key: model.maxSessionTurns

Wall-clock budget for the run; budget aborts exit with code 55.

accepts 90 (s), 30s, 5m, 1h, 1.5h

Cumulative tool-call budget for the run; budget aborts exit with code 55.

Settings key: model.maxToolCalls

Write structured JSON events to file descriptor n; mutually exclusive with --json-file.

number, n >= 3

Write structured JSON events to a file; the path can be a regular file, a FIFO, or /dev/fd/N.

path

Watch this file for JSONL commands written by an external program.

path

Option of qwen mcp add: Configuration scope (user or project).

default user

Option of qwen mcp add: Transport type (stdio, sse, http).

default stdio

Option of qwen mcp add: Set environment variables.

Option of qwen mcp add: Set HTTP headers for SSE and HTTP transports.

Option of qwen mcp add: Set connection timeout in milliseconds.

Option of qwen mcp add: Trust the server; skip confirmations in trusted workspaces.

default false

Option of qwen mcp add: Set the description for the server.

Option of qwen mcp add: A comma-separated list of tools to include.

default all tools included

Option of qwen mcp add: A comma-separated list of tools to exclude.

default none

Option of qwen mcp add: OAuth client ID for MCP server authentication (sse/http only).

Option of qwen mcp add: OAuth client secret for MCP server authentication.

Option of qwen mcp add: OAuth redirect URI for authentication callback.

default http://localhost:7777/oauth/callback

Option of qwen mcp add: OAuth authorization URL.

Option of qwen mcp add: OAuth token URL.

Option of qwen mcp add: OAuth scopes (comma-separated).

Option of qwen serve: TCP port. 0 = OS-assigned ephemeral port.

default 4170

Option of qwen serve: Bind interface; anything beyond loopback requires a token.

default 127.0.0.1

Option of qwen serve: Bearer token. Falls back to the QWEN_SERVER_TOKEN env var.

Option of qwen serve: Absolute workspace directory registered by this daemon; repeatable.

default process.cwd()

Start the whole session inside a git worktree; cannot be combined with --acp/--experimental-acp.

Constrain the model's final answer to a supplied JSON Schema; headless only.

Bare mode: ignores most settings-derived inputs and turns off all hooks (or QWEN_CODE_SIMPLE).

Disable the named built-in tools for the run (comma-separated).

Comma-separated tool names

Allowlist of core tools (equivalent of tools.core); structured_output bypasses it.

Allowlist of MCP server names for the run; when set, mcp.excluded is ignored.

Selects the auth type/protocol at launch; combine with --model to point at a provider entry.

API key for the active OpenAI-compatible provider; highest-priority credential source.

Base URL for the active OpenAI-compatible provider.

Deprecated flag replaced by --acp (Agent Client Protocol mode).

Fallback model IDs to try on capacity errors (429/503/529); CLI form of modelFallbacks.

Supplies MCP servers for the current invocation (path-taking flag).

File to write telemetry to (path-taking flag).

Skip TLS certificate verification (or QWEN_TLS_INSECURE=1); only for trusted networks.

List recent conversation sessions with metadata.

--json (default false), --limit

List interactive sessions running right now.

--json

Manage trusted controller tokens.

Add an MCP server configuration (stdio, sse or http).

qwen mcp add [options] <name> <commandOrUrl> [args...]

Remove an MCP server configuration by name.

List configured MCP servers.

Install an extension from a marketplace, git repo, npm package, local path or archive URL.

--scope project|workspace, --registry

Uninstall an installed extension.

Disable an extension at the user level, or for one workspace with --scope=workspace.

Enable an extension, optionally for a specific workspace with --scope=workspace.

Update an installed extension, or all of them with --all.

Link a local extension directory, for example an Agent Plugin under development.

Create a new extension scaffold.

Add a marketplace source (owner/repo, git URL, https URL to marketplace.json, or local path).

List configured marketplace sources.

Re-fetch a marketplace's plugin listing.

Remove a marketplace source.

Set an extension setting value.

[--scope user|workspace]

List an extension's settings.

Start all configured channels in a shared agent process, or a single named channel.

Gracefully stop the running channel service (or the daemon selection with --daemon-url).

--daemon-url, --token

Show whether the channel service is running, its uptime and session counts per channel.

--daemon-url, --token

Select channels on a token-protected qwen serve daemon at runtime.

List pending pairing requests for a channel.

Approve a pairing request by its code.

Configure the WeChat (Weixin) channel.

Start the experimental HTTP daemon serving the API and the built-in Web Shell (default http://127.0.0.1:4170/).

Update a standalone installation; honors QWEN_UPDATE_BASE_URL.

Report the active Linux tool execution sandbox boundary, verify it with --verify, or run a command inside it.

--verify

Headless wrapper that runs /review non-interactively; exit 0 completed, 1 no verdict, 3 blocking.

[target] [--json] [--fail-on request-changes] [--comment] [--resume] [--quiet]

Verify Batch API setup; nothing is billed.

Submit a batch plan; --dry-run previews it, --expect <snapshot> submits exactly that snapshot.

--dry-run, --expect

Validate results and write target files, optionally waiting for the batch to finish.

Resubmit only the failed items; truncated items need a larger --max-output-tokens.

List every recorded batch task with its project.

Cancel a batch task; partial results are still billed.

Delete the local record of a batch task (cancels nothing); --force overrides the safety refusal.

Create a task on an Agent Board and print its id (experimental).

--board, --as

Show a board's tasks and asks; --as filters to that actor, --json emits JSON.

Claim a board task.

Complete a claimed board task, optionally with --note.

Ask another actor a question; with --wait exit 0 answered, 2 declined, 3 TTL expired, 4 wait ended.

--wait, --timeout, --ttl

Answer an ask on the board.

Decline an ask on the board.

Remove settled records older than a cutoff in days.

--older-than <days>

Removed. Legacy invocations print a removal notice with migration guidance; use /auth instead.

Close dialogs and suggestions; with an empty prompt, cancel an ongoing request.

Cancel the ongoing request and clear the input. Press twice to exit the application.

Exit the application if the input is empty. Press twice to confirm.

Clear the screen.

Toggle expanded detail mode: expand or collapse all thinking blocks and tool outputs inline.

Stashes non-empty input for the current project and restores it on the next launch.

Toggle the display of tool descriptions.

Toggle Markdown output between rich rendered previews and raw/source mode.

Cycle approval modes (plan, default, auto-edit, auto, yolo).

Toggle shell mode when the input is empty.

Toggle keyboard shortcuts display when the input is empty.

Open slash-command completion.

Open completion for files, folders, and other context.

Start voice dictation when it and a voice model are configured.

Insert a newline.

Submit the current prompt. While a response is running, steer the current turn.

Queue the current prompt or command for the next turn instead of steering.

When queued messages are present, move them back into the input for editing.

Autocomplete the current suggestion if one exists.

Clear the input prompt.

Reverse search through input/shell history.

Retry the last failed request.

Paste clipboard content; an image is saved and a reference to it is inserted in the prompt.

Open the current input in an external editor.

Undo the last input edit.

Redo the last undone input edit.

Delete from the cursor to the end of the line.

Delete from the cursor to the beginning of the line.

Toggle keyboard focus between the running foreground shell and the prompt.

Scroll the focused shell up or down.

Promote the foreground shell to a background task.

Scroll history up one line (when ui.useTerminalBuffer is enabled).

Scroll history down one line.

Jump to the top of the conversation.

Jump to the bottom (and re-engage live auto-follow).

See context CLI received from IDE.

The preferred editor to open files in.

string; default undefined

Name of the output style (built-in Concise, Proactive, Explanatory, Learning, or a custom style).

string; default undefined

Enable Vim keybindings.

boolean; default false

Enable automatic update checks and installations on startup.

boolean; default true

Auto-show a one-line "where you left off" recap when returning to the terminal.

boolean; default false

Add a Co-authored-by trailer and a per-file AI-attribution git note to commits made through Qwen Code.

boolean; default true

Append a Qwen Code attribution line to pull request descriptions when running gh pr create.

boolean; default true

Language for the user interface; "auto" detects from system settings. Requires restart.

enum; default "auto"

Language for model output; "auto" detects from system settings. Requires restart.

string; default "auto"

Save chat history to disk. Disabling this also prevents --continue and --resume from working.

boolean; default true

The format of the CLI output.

string; values "text", "json", "stream-json"; default "text"

Default effort for /review when neither --effort nor a remembered level applies.

enum: low, medium, high, auto; default "auto"

The color theme for the UI.

string; default "Qwen Dark"

Custom theme definitions.

object; default {}

Custom status line configuration (supports command, refreshInterval, respectUserColors, hideContextIndicator).

object; default undefined

Hide the window title bar.

boolean; default false

Hide all tips (startup and post-response) in the UI.

boolean; default false

Hide the startup ASCII logo and info panel.

boolean; default false

Show line numbers in code blocks in the CLI output.

boolean; default true

Default Markdown display mode: "render" for rich previews or "raw" for source-oriented Markdown.

string; default "render"

Show welcome back dialog when returning to a project with conversation history.

boolean; default true

Enables screen reader mode, which adjusts the TUI for better compatibility with screen readers.

boolean; default false

Enable followup suggestions that predict what you want to type next after the model responds.

boolean; default true

Render conversation history in an in-app scrollable viewport instead of the terminal scrollback buffer.

boolean; default true

Enable in-app SGR mouse tracking.

boolean; default true

Enable IDE integration mode.

boolean; default false

Enable collection of usage statistics.

boolean; default true

The Qwen model to use for conversations.

string; default undefined

How hard reasoning-capable models think, applied across all providers; set with /effort.

enum: low, medium, high, xhigh, max; default undefined

Maximum recorded prompt token count allowed before sending the next message; -1 means unlimited.

number; default -1

Maximum number of user/model/tool turns to keep in a session. -1 means unlimited.

integer; default -1

Wall-clock budget for headless / unattended runs, in seconds; overridable via --max-wall-time.

number; default -1

Cumulative tool-call budget for a run; -1 unlimited, 0 no tool calls. Aborts with exit code 55.

number; default -1

Maximum sub-agent nesting depth (1-based levels); values clamp to 1-100.

number; default 5

Advanced overrides passed to the underlying content generator (timeout, maxRetries, samplingParams, etc.).

object; default undefined

Enables logging of OpenAI API calls for debugging and analysis.

boolean; default false

Custom directory path for OpenAI API logs; defaults to logs/openai.

string; default undefined

Model used for generating prompt suggestions and speculative execution; empty uses the main model.

string; default ""

Model used by /advisor for second-opinion reviews of the conversation.

string; default ""

Image-capable model used as the vision bridge when a text-only main model receives an image.

string; default ""

Model used for chat compression (auto-compaction). Leave empty to fall back to the main model.

string; default ""

Model used by the built-in image_gen tool.

string; default ""

Model used for voice transcription; voice dictation stays disabled until one is selected.

string; default ""

Ordered fallback model IDs (comma-separated, max 3) tried on capacity errors (429/503/529).

string; default ""

Optional per-model pricing for cost estimation in /stats model.

object; default undefined

The name of the context file(s).

string or array of strings; default undefined

Target fraction of the context window at which auto-compaction triggers.

number; default undefined (uses internal 0.85)

Additional directories to include in the workspace context.

array; default []

Controls whether /memory refresh loads QWEN.md files from all added directories.

boolean; default false

Respect .gitignore files when searching.

boolean; default true

Respect .qwenignore and configured custom ignore files when searching.

boolean; default true

When true, enables fuzzy search capabilities when searching for files.

boolean; default true

Sandbox execution environment (can be a boolean or a path string).

boolean or string; default undefined

Sandbox image URI used by Docker/Podman when --sandbox-image and QWEN_SANDBOX_IMAGE are unset.

string; default undefined

Use node-pty for an interactive shell experience.

boolean; default mode-based

Deprecated. Will be removed in next version.

array of strings; default undefined

Deprecated. Use permissions.deny instead.

array of strings; default undefined

Deprecated. Use permissions.allow instead.

array of strings; default undefined

Sets the default approval mode for tool usage.

string; default auto

Command to run for tool discovery.

string; default undefined

Command to run for calling a discovered tool.

string; default undefined

Use ripgrep for file content search instead of the fallback implementation.

boolean; default true

Enable the Workflow tool, which lets the model author and run scripts that orchestrate subagents.

boolean; default false

Truncate tool output if it is larger than this many characters.

number; default 25000

Enable background extraction of memories from conversations.

boolean; default true

Enable automatic consolidation (deduplication and cleanup) of collected memories.

boolean; default true

Enable a project memory tier shared with collaborators via the git-tracked .qwen/team-memory/ directory.

boolean; default false

Rules for auto-approved tool calls (no confirmation needed).

array of strings; default undefined

Rules for tool calls that always require user confirmation. Takes priority over allow.

array of strings; default undefined

Rules for blocked tool calls. Highest priority - overrides both allow and ask.

array of strings; default undefined

Slash command names to hide and refuse to execute.

array of strings; default undefined

Hard-disabled skill names.

array of strings; default undefined

Explicit skill opt-ins; cancels a matching skills.defaultDisabled entry.

array of strings; default undefined

Command to start an MCP server.

string; default undefined

An allowlist of MCP server names to make available to the model; supports glob patterns.

array of strings; default undefined

A denylist of MCP servers to exclude; a server in both lists is excluded.

array of strings; default undefined

Idle timeout in milliseconds for MCP tool calls (10000 to 3600000).

number; default 300000

Setting to track whether Folder trust is enabled.

boolean; default false

The currently selected authentication type.

string; default undefined

The required auth type (useful for enterprises).

string; default undefined

Environment variables to exclude from project context.

array of strings; default ["DEBUG","DEBUG_MODE"]

Configuration for the bug report command; overrides the default URL for /bug.

object; default undefined

Custom directory for approved Plan Mode files (top-level key).

string; default undefined

Enable in-session cron/loop tools (cron_create, cron_list, cron_delete). Requires restart.

boolean; default true

Enable agent-team collaboration tools for multi-agent coordination. Requires restart.

boolean; default false

The command to execute to start the MCP server via standard I/O.

string; optional

Arguments to pass to the command.

array of strings; optional

Environment variables to set for the server process.

object; optional

The working directory in which to start the server.

string; optional

The URL of an MCP server that uses Server-Sent Events (SSE) for communication.

string; optional

The URL of an MCP server that uses streamable HTTP for communication.

string; optional

A map of HTTP headers to send with requests to url or httpUrl.

object; optional

Timeout in milliseconds for requests to this MCP server.

number; optional

Trust this server and bypass its tool call confirmations in a trusted workspace.

boolean; optional

List of tool names to include from this MCP server (allowlist behavior).

array of strings; optional

List of tool names to exclude from this MCP server.

array of strings; optional

Whether or not telemetry is enabled.

boolean

The endpoint for the OTLP Exporter.

string

The protocol for the OTLP Exporter (grpc or http).

string

Whether or not to include user prompt content and API request/response text in the logs.

boolean

Path to write telemetry to a file. When set, overrides OTLP export.

string

Declares which models are available and how to connect to them; keys are API protocols (openai, anthropic, gemini).

Stores API keys directly in settings.json as a lowest-priority fallback (shell export and .env take precedence).

Customizes the global configuration directory (default: ~/.qwen).

Overrides the runtime output directory (conversations, logs, todos); defaults to QWEN_HOME.

Set to true or 1 to enable usage statistics; any other value disables them.

Overrides the privacy.usageStatisticsEnabled setting.

Set to true or 1 to enable telemetry; any other value disables it.

Overrides the telemetry.enabled setting.

Sets the OTLP endpoint for telemetry.

Overrides the telemetry.otlpEndpoint setting.

Sets the file path to write telemetry to. When set, overrides OTLP export.

Overrides the telemetry.outfile setting.

Alternative to the sandbox setting in settings.json; overrides the CLI flag and settings.

Accepts true, false, docker, podman, or a custom command string

Overrides sandbox image selection for Docker/Podman.

Takes precedence over tools.sandboxImage

(macOS specific) Switches the Seatbelt (sandbox-exec) profile on macOS.

default permissive-open

Set to true or 1 to enable verbose debug logging; excluded from project .env files by default.

Set to true or 1 to enable verbose debug logging; excluded from project .env files by default.

Set to any value to disable all color output in the CLI.

Override the OSC 8 clickable-link detection in the markdown renderer.

1 to force-enable; 0 to disable

Set to 1 to hard-disable OSC 8 clickable hyperlinks in the markdown renderer.

1

Set to a string to customize the title of the CLI.

Overrides the default maximum output tokens per response; disables automatic escalation.

Set to true or 1 to retry transient 429/529 capacity errors indefinitely with backoff.

Set to 1 to write a JSON startup timing report to ~/.qwen/startup-perf/.

Set to 1 to wait synchronously for every MCP server's discover handshake at startup.

Set to off to ignore the bundled models.dev model catalog.

off

Set to off to stop the daily models.dev download into ~/.qwen/model-registry.json.

off

API key for the OpenAI-compatible protocol; the default envKey for openai entries.

Base URL for the OpenAI-compatible protocol.

Model ID for the OpenAI-compatible protocol (alias: QWEN_MODEL).

Alias of OPENAI_MODEL for the OpenAI-compatible protocol.

API key for the Anthropic protocol (modelProviders key anthropic).

Base URL for the Anthropic protocol.

Model ID for the Anthropic protocol.

API key for the Google GenAI protocol (modelProviders key gemini).

Model ID for the Google GenAI protocol.

API key for Vertex AI; used with GOOGLE_MODEL and sets GOOGLE_GENAI_USE_VERTEXAI=true.

Model ID for Vertex AI entries.

Google Cloud project for keyless Vertex AI access via Application Default Credentials.

Alibaba Cloud DashScope API key, commonly used as the envKey of Qwen model provider entries.

API key for QwenCloud used in headless or scripted setup.

API key for the Alibaba Cloud Coding Plan in headless or scripted setup.

OpenRouter API key; used with OPENAI_BASE_URL=https://openrouter.ai/api/v1.

Overrides the path of the system settings file (highest-precedence settings file).

Overrides the path of the system defaults settings file (lowest-precedence settings file).

Custom HTTPS release root for standalone updates (qwen update, /update, automatic updates).

HTTPS URL without credentials, query string or fragment

Denylist of slash commands; unioned with slashCommands.disabled and --disabled-slash-commands.

Comma-separated command names

Set to 1 to force the Workflow tool on (overrides tools.workflowsEnabled).

1

Set to 1 to enable agent-team collaboration tools (same as experimental.agentTeam).

1

Set to 1 to disable the in-session cron/loop tools (experimental.cron).

1

Sets the per-request API timeout in milliseconds; 0 disables the request timeout.

default 120000

Global idle timeout between streamed chunks; 0 disables the idle guard.

default 240000

Equivalent of --safe-mode: disables all customizations to isolate problems.

true

Equivalent of --bare; among other effects it turns off all hooks.

Set to 1 to suppress the stderr warning when --yolo runs headless without a sandbox.

1

Set to the project directory in the environment of every command hook.

Highest-priority source for UI language auto-detection.

Set to 1 to skip TLS certificate verification (same as --insecure).

1

Injects custom flags into the docker or podman command for container-based sandboxing.

Space-separated flags

On Linux, forces (true) or disables (false) UID/GID mapping so the sandbox runs as your user.

true or false

Bearer token for qwen serve when --token is not passed.

Controls Mermaid image rendering in the terminal markdown renderer.

Documented value: 1

User settings file; applies to all Qwen Code sessions for the current user.

Project settings file in the project root; overrides user settings for that project.

System defaults file (Linux path); lowest-precedence settings layer. Override with QWEN_CODE_SYSTEM_DEFAULTS_PATH.

Windows: C:\ProgramData\qwen-code\system-defaults.json; macOS: /Library/Application Support/QwenCode/system-defaults.json

System settings file (Linux path); overrides user and project settings. Override with QWEN_CODE_SYSTEM_SETTINGS_PATH.

Windows: C:\ProgramData\qwen-code\settings.json; macOS: /Library/Application Support/QwenCode/settings.json

Recommended project env file for Qwen Code variables; never filtered by the default excluded-vars list.

Environment file loaded automatically; DEBUG and DEBUG_MODE are excluded from project .env files by default.

Project-specific history of shell commands run in the CLI.

Default context file for hierarchical instructional context (configurable via context.fileName); /init creates one.

Global context file that applies to all your projects.

Personal, project-local context file.

Default output of /summary; read by the welcome-back flow.

Ignore file whose patterns exclude files from Qwen Code tools.

Custom macOS Seatbelt profile selected with SEATBELT_PROFILE=<profile_name>.

Custom sandbox image definition for container-based sandboxing.

Project Agent Skill; each Skill is a directory containing a SKILL.md.

Global custom slash command files (Markdown recommended; TOML deprecated); test.md becomes /test.

Project custom slash command files; subdirectories form namespaces such as /git:commit.

User-level custom output style (project styles go in <project>/.qwen/output-styles/).

User-level rules directory (or $QWEN_HOME/rules/); project rules go in <project>/.qwen/rules/.

LSP server configuration file in the project root (LSP is experimental, enabled with --experimental-lsp).

Default MCP OAuth token store (plaintext, mode 0600).

Extension manifest file; Claude and Gemini manifests are converted to it on install.