Configure options below — the preview updates in real time.
Top-level configuration including policy paths.
Additional security or operational policy files/directories to load.
Admin-enforced policy files (Enterprise override).
Privacy and usage statistics settings.
Editor preferences, update behavior, approval mode, and session settings.
The preferred editor executable to open files in (e.g., "code", "vim").
Default approval mode for tool execution. "default" prompts user; "auto_edit" auto-approves code edits; "plan" is read-only.
Maximum number of network retry attempts for requests to the main chat model.
Auto-delete chats older than this period (e.g. "30d").
Minimum retention period safety limit (e.g. "1d").
Configure planning artifacts directory and model routing.
The directory path where intermediate planning artifacts are stored.
Output format settings.
The output format for CLI responses.
Configure the Gemini model, context history, and loop prevention.
The Gemini LLM to use for conversations.
Max user/model/tool turns to keep in a session context window. -1 is unlimited.
The fraction of context capacity at which to trigger context compression (e.g., 0.5 = 50%).
Configure context file names, directories, and search behavior.
Target file names containing hierarchical instructions/memory.
Additional workspace context directories.
Maximum directories to search for memory.
Paths to additional ignore files beyond .gitignore (earlier takes precedence).
Control visual output including thinking display, warnings, footer elements, and accessibility.
The color theme for the CLI UI.
Controls whether the "chain of thought" output is visible inline.
Loading display mode.
Error display verbosity level.
Tool execution, sandboxing, shell, and discovery settings.
Legacy full-process sandbox execution environment.
Pager command for shell output.
Max seconds without shell output before timeout.
Max characters for truncated tool outputs (0 disables).
Authentication, YOLO mode, folder trust, environment variable redaction, and extension security settings.
The currently selected authentication type.
Required authentication type (enterprise).
Environment variables to always allow (bypass redaction).
Environment variables to always redact.
Regex pattern list. If populated, only extensions matching these patterns are executed.
Experimental features for tool output masking, plan mode, clipboard handling, and agents.
Billing and overage settings.
Strategy for handling billing overages.
IDE integration settings.
Advanced settings for memory, DNS, and environment variables.
DNS resolution order.
Environment variables to exclude from context.
Browser agent and agent override settings.
Browser session mode.
Agent skills settings.
List of disabled skill names.
Hooks system settings.
Model Context Protocol server definitions for external agent tools.
Allowlist of MCP servers. Only servers named here will be permitted to start.
Denylist of MCP servers to exclude.
Telemetry and observability settings.
Telemetry destination.
OTLP Exporter endpoint URL.
OTLP protocol.
File path for local telemetry output.
1