Skip to content

Settings file reference

The 307 keys Belai reads from its settings files, by group.

Keys Belai reads from its settings files, grouped by the part of the configuration they belong to. The guide pages explain what each group does; this page is the full list of names.

bashrewrite

Explained in Built-in tools.

enabled, match, replace, rules

budgets

Explained in Token budgets and session intelligence.

model, plan_limits, provider, scope, tokens

codemode

Explained in Built-in tools.

enabled, max_calls, max_output_bytes, model, provider, tier, timeout_ms

firewall

Explained in Package and AI firewalls on factory work.

active, adapter, enabled, header, instances, mode, permission_auto_apply, permission_offers, providers, url

gitrepos

Explained in The library: skills, prompts, commands, processes, repos and more.

auth, depth, dir, enabled, installation_id, kind, name, refs, submodules, url, visibility

jev

Explained in Guardrails, classifier, sandbox and trust.

jobs, locate_previews, thresholds

jevthresholds

Explained in Guardrails, classifier, sandbox and trust.

align_at, allow_at, clear_at, cover_at, deny_at, drop_at, goal_complete_at, goal_not_started_at, goal_rival_max, hit_at, keep_at, lead_at, mode_confident, mode_headless, mode_margin, option_hit, option_lead, option_margin, route_at, simple_at, strong_at, swap_at, topic_at, triage_at, voice_at

knowledge

Explained in Knowledge.

max_index_tokens, max_project_tokens, max_result_tokens, topic_budget_docs, topic_chunks

lsp

Explained in Language servers.

classify_diagnostics, enabled, fallback, languages, max_diagnostics, max_repair_attempts, servers, timeout_ms

mcp

Explained in MCP servers.

clef, enabled, skip_ask, skip_ask_at, vulnetix

offload

Explained in Sessions: resume, export, compact and rename.

enabled, preview_tokens, threshold_tokens

permissions

Explained in Agent questions, answers and permission offers.

allow, ask, block, deny

projectprefs

Explained in Git sync and repository info.

agent, ask_permission, auto_commit_per_task, caveman, enabled, firewall_enabled, guardrails, jev, jobs, languages, layout, lsp, mode, on_fail, post_end, show_edits, show_internal_work, show_reasoning, show_tool_calls, tests, thresholds

screenshot

desktop, enabled

settings

a2a, agents, allow_project_providers, allow_project_workspace_dirs, allowed_programs, api, api_key_env, args, ask_permission, auth, auto_commit_per_task, autofix, backend, banner, base_url, bash_readonly, bash_rewrite, batch, budget_cycle_seconds, budget_warn, budgets, builtin, caches, caveman, chunk, classifier, clipboard_images, clone, code, code_scan, colors, command, commands, concurrency, context_window, context_windows, continue_on_error, decision, decision_path, defer_tools, dep_watch, disabled, effort, enabled, env, events, extra_writable, fast_model, filter, firewall, firewall_enabled, forges, gateway_url, git, goal_explore, guardrails, headers, hooks, host, id, idle_pane, images, intel, jev, kanban, kind, kitty_keyboard, knowledge, layout, lsp, mac_key_hints, max_agents, max_attempts, max_background_processes, max_bytes, max_clarify_rounds, max_explore_iterations, max_fix_passes, max_iterations, max_passes, max_process_recoveries, max_state_bytes, max_tokens, max_workers, mcp, mcps, metrics, min_turn_seconds, mode, mode_detection, model, models, mouse, name, network, notifications, offload, on_fail, org_id, otlp_endpoint, permissions, phase1, phase2, plan_explore, port, post_end, processes, profiles, prompts, protocol, provider, provider_labels, providers, publish, push, read_only, remote_answers, remote_prompts, report, repos, resilience, rewrites, routing, sandbox, scope, screenshot, secrets, self_authoring, servers, session_retention_days, show_a2a, show_edits, show_internal_work, show_reasoning, show_session_names, show_todos, show_tool_calls, skills, source, spinner, status_bar, subcommands, sync, telemetry, teleport, tests, threshold, tier, timeout, timeout_ms, timeout_seconds, token_budgets, tools, traces, transport, tts, ui, update_check, url, use_cases, voice, vulnetix, vulnetix_sweep_enabled, vulnetix_sweep_roots, web_fetch, workspace_dirs

state

active_goal, active_plan, active_profile, active_session, effort, last_mode, model, onboarded_at, provider

tts

cache_mb, consented, enabled, read_reports, speed, voice

voice

cleanup, commands, delivery, device, enabled, key, log, mode, wake_word

webfetch

cache, cache_ttl_seconds, index, index_tokens

esc
↑ ↓ move↵ openSearches in your browser. A key like autonomy.max or a code like threshold_required jumps straight to it.