Settings

Launching

ait settings

The Settings TUI requires the shared Python virtual environment (installed by ait setup) with the textual and pyyaml packages.

Understanding the Layout

The Settings TUI organizes configuration into five tabs, each accessible via a keyboard shortcut displayed in the footer.

Agent Defaults (a)

×

Shows the default agent/model for each operation (task-pick, explain, batch-review, raw). These defaults are used when launching tasks from the Board TUI and running explain from the Code Browser TUI. Each entry displays:

• The current value (e.g., claudecode/opus4_6)
• A layer badge: [PROJECT] (green) for project-level defaults or [USER] (amber) for per-user overrides
• Verified score context – score, run count, and recency (e.g., [96 (9 runs, 2 this mo)])
• When the same underlying LLM has scores from multiple providers, an all providers summary appears below (e.g., all providers: 96 (12 runs, 3 this mo))
• A description of what the operation does

Press Enter on any operation to change its agent/model through the model picker.

The picker opens with a Top Verified list showing the highest-scoring models for the selected operation across all providers. Select a model directly or choose Browse all models to use the full agent/model browser.

Press d or Delete on an entry to remove a user override and revert to the project default.

Board (b)

Displays board configuration in two sections:

• Columns (read-only) – Lists each board column’s ID, title, and color
• User Settings (editable) – Auto-refresh (min) (cycle through 0, 1, 2, 5, 10, 15, 30 where 0 = disabled) and Sync on refresh (toggle whether push/pull runs on each auto-refresh)

Press Save Board Settings to persist changes.

Project Config (c)

Edit shared values from aitasks/metadata/project_config.yaml directly in the TUI. The initial editable keys are:

• codeagent_coauthor_domain – Shared email domain used for custom code-agent Co-authored-by trailers
• verify_build – Build verification command or YAML list of commands run after implementation

Press Enter on a row to edit the value, then use Save Project Config to persist it.

Models (m)

Read-only display of all available models organized by agent (claudecode, codex, geminicli, opencode). For each model, shows:

• Name – Internal identifier used in agent strings
• CLI ID – Exact model ID passed to the CLI
• Notes – Description of the model
• Verified stats – Per-operation score with run count and recency (e.g., pick: 96 (9 runs, 2 this month))
• All providers – When the same LLM is available through multiple providers, a cross-provider aggregate line appears below the model row

Model definitions are managed by editing aitasks/metadata/models_<agent>.json files directly or using the /aitask-refresh-code-models skill.

Profiles (p)

Browse and edit execution profiles – YAML files that pre-answer workflow questions to reduce interactive prompts. Each profile contains settings organized into groups:

• Identity – name, description
• Task Selection – skip_task_confirmation, default_email
• Branch & Worktree – create_worktree, base_branch
• Planning – plan_preference, plan_preference_child, post_plan_action
• Feedback – enableFeedbackQuestions
• Exploration – explore_auto_continue
• QA – qa_mode, qa_run_tests
• Lock Management – force_unlock_stale
• Remote Workflow – done_task_action, orphan_parent_action, complexity_action, review_action, issue_action, abort_plan_action, abort_revert_status

Fields use type-appropriate controls: boolean keys toggle between true/false/(unset), enum keys cycle through their options, and string keys open an edit dialog. Setting a value to (unset) removes it from the profile so the question is asked interactively at runtime.

Navigating

Next: How-To Guides — common tasks like editing agents, creating profiles, exporting bundles. Or jump to the Reference for the full tab/field/profile schemas.