Core:
talon runtalon continue --last|--session <id> [task]talon tuitalon tui --continuetalon tui --resume <session_id|title>talon session-api servetalon opstalon dashboard(compatibility alias oftalon ops)talon inittalon doctortalon version
Operational:
talon task list|show|timelinetalon session list|show|archive|search <query>talon session handoff --session <id> --adapter <adapter> [--external-session <id>]talon schedule create|list|show|edit|pause|resume|run-now|runs|remove|status|tick|runtalon inbox [--status <status>]talon inbox list|show|done|dismisstalon commitments list|show|create|block|unblock|complete|canceltalon next list|add|done|block|unblock|resumetalon trace [task_id] [--summary]talon trace context <task_id>talon audit <task_id> [--summary]talon approve pending|allow|denytalon workspace map|rollbacktalon repo map(deprecated alias)
Subsystems:
talon provider list|current|status|setup|use|promote|test|smoke|stats|routetalon provider alias list|add|removetalon provider custom list|add|removetalon provider credential list|add-env|disable|enable|removetalon provider fallback list|add|remove|clear [--slot <slot>]talon model|model list [--json] [--session <id>]|model status [--json] [--session <id>]|model set <selection> [--workspace|--session <id>]|model clear --session <id>talon model auxiliary list|set|resettalon budget show --task <id>|--session <id>talon memory list|show|search|snapshot|review|guide|add|forget|why|review-queuetalon experience list|show|review|promote|searchtalon skills list|view|enable|disable|draft|promote|rollbacktalon gateway serve-webhook|serve-feishu|list-adapterstalon mcp list|ping|servetalon sandboxtalon config doctor
Maintainer diagnostics for source checkouts:
talon replay <task_id> [--dry-run]talon eval run --fixture <path> [--explain]talon eval smoke --fixture <path>(compatibility alias)talon smoke run --fixture <path>talon eval betatalon release check(maintainer-only; run from the auto-talon repository root)
TUI slash commands (chat mode):
/clear [name](save current session and start a new one)/new [title](start a fresh named assistant session)/branch [name](fork current transcript)/handoff <adapter> [external-session-id]/handoff status/sessions(session picker)/resume <session-id-prefix|title>/modelor/model list(show current model source and numbered configured models)/model status(show aliases, fallback chain, auxiliary slots, and env-only selections)/model 1(switch the active session by configured model number)/model <provider:model>(set a strict model override for this session without restarting TUI)/model default(clear the current session model override)/model <provider:model> --global(switch and save to user provider config)/model <provider:model> --workspace(switch and save to workspace provider config)/model <alias>(usemodelAliasesfrom provider config)- Session model overrides take priority over
routing.providersuntil cleared - Auxiliary slots set to
autofollow the current main provider after a/modelswitch - Alias selections persist as the resolved provider name in
provider.config.json /next list [session-id-prefix]/next done <next-action-id-prefix>/next block <next-action-id-prefix> <reason...>/commitments list [session-id-prefix]/commitments done <commitment-id-prefix>/commitments block <commitment-id-prefix> <reason...>/schedule list [active|paused|completed|archived|all]/schedule create <when> | <prompt>/schedule pause|resume|run-now|runs|remove <schedule-id-prefix>/memory/memory review/memory add <profile|project> <text>/memory forget <memory-id-prefix>/memory why [memory-id-prefix]
Schedule notes:
talon schedule createaccepts--at 30m,--every 2h, and--execution-mode isolated|continue|session:<id>.--sessionrequires an execution mode; isolated schedules always run in a fresh session. Advanced fields (deliveryTargets,skills,toolsets,noAgent,repeatRemaining) are available via the agentcronjobtool or gateway API.talon schedule preview <expr> [--timezone TZ] [--count N]validates one timing expression and prints upcoming fire times without creating a schedule.talon schedule run <schedule-id> --wait [--timeout MS] [--poll-interval MS]blocks until the run reaches a terminal state.- Agent
cronjobtool manages schedules in chat; scheduled runs cannot nest further schedule edits unless you opt in. - Scheduled agent runs are scanned for prompt-injection patterns before execution; set
metadata.allowDelegate: trueto allowdelegate_taskduring a scheduled run. - Feishu
/schedule create cron <expr> | <prompt>and/schedule edit <id> cron <expr>expose native cron in chat; unsupported timing or missing toolsets fail explicitly instead of falling back to another execution path.
Search tool notes:
- The runtime
search_filestool is a grep-style workspace search tool. It supports literal or regex queries, include/exclude globs, filename matching, and result modes:matches,files, andcount. matchesreturns line hits with context,fileslists matching files, andcountreports per-file and total match counts.search_filesuses ripgrep for content matches when available, with a Node implementation as the only compatibility path. It does not add semantic search, vector recall, or LSP/code-intelligence behavior.