Docs v1.2.3

Commands

Type / to open command search. TatsuCode commands control models, providers, sessions, display, permissions, and workflow features.

Typing a partial command shows matching commands with a description; Tab completes, Enter runs


Quick Reference

CommandWhat It Does
/newStart a new session (optionally switch project)
/clearClear visible conversation output
/modelsOpen model selector
/models-addAdd models to your available list
/models-taskagentsAssign cost-efficient models to Task Agents
/connectOpen provider connections
/providersAlias of /connect
/skillsBrowse and load skills
/skills-refreshRe-scan skill directories
/pluginsView and manage plugin connections
/plugins-refreshRe-scan plugin directories
/usageShow token usage
/usage-quotaShow provider quotas/rate limits
/compactManage context compaction
/sessionOpen saved sessions
/session-renameRename current session
/projectsSwitch between projects
/themeChange UI theme
/soundsConfigure notification chimes
/permissionsManage tool permissions
/yoloToggle YOLO mode
/image-generationConfigure image generation
/image-galleryOpen image gallery
/temperature / /tempSet temperature
/reason-effortSet reasoning effort
/reason-displayToggle reasoning visibility
/diff-viewDiff display mode
/verbosity(OpenAI Plus/Pro) Response verbosity
/image-detail(OpenAI Plus/Pro) Image detail level
/native-search(OpenAI Plus/Pro) Toggle native web search
/initCreate / update AGENTS.md
/add-tatsuAdd to system PATH
/remove-tatsuRemove from PATH and config
/exit / /quitExit TatsuCode

Session & Project Commands

/new

/new
/new C:\Projects\my-app
/new ~/projects/backend-api

Starts a fresh session. If a path is provided, switches project too.

/session

Opens saved sessions for the active project.

/session-rename

/session-rename Feature: OAuth migration

Gives your current session a readable name.

/projects

Shows recent projects and lets you switch quickly.


Model & Provider Commands

/models

Opens model picker with filtering and provider grouping.

/models-add

Add models to your available list.

/models-taskagents

Assign a different model to each parallel sub-agent (Web, Code Explorer, Design Reviewer / Architect, YouTube Video, PDF Reader). Use cheaper / faster models for scout work and reserve stronger models for the main chat and review/design work. See Task Agents.

/connect (primary)

Open Provider Connections:

  • OAuth subscriptions (OpenAI Plus/Pro, Claude Agent / Claude Code)
  • Decision models (TypeSafe AI key)
  • OpenRouter API key
  • Custom/local providers
  • Web search providers

/providers (alias)

Kept for compatibility; routes to the same provider UI as /connect.


Agent Skills & Plugins Commands

/skills

Browse and load skills into the current session.

/skills-refresh

Force skill re-discovery after editing or adding skill files.

/plugins

View and manage plugin connections: Unity, Unreal Engine, Godot, Blender, and Autodesk Maya. See Game Engine Plugins.

/plugins-refresh

Re-scan plugin directories after install or update.


AI Behavior & Display Commands

/temperature or /temp

/temp 0.3
/temperature 0.7

Adjust response creativity.

/reason-effort

Adjust reasoning depth on supported models.

/reason-display

Toggle display of reasoning/scratch output.

/diff-view

Switch diff display mode (unified or side-by-side).

/theme

Open theme selector.

/sounds

Configure notification chimes for task-complete and permission/question events. Plays only when the TatsuCode window is unfocused. See Notifications.


Image Commands

/image-generation

Configure image-generation provider, model, and defaults (size, aspect ratio, quality, background). See Images.

/image-gallery

Open the unified image gallery — every image in your conversation and project, in one searchable view.


OpenAI Plus/Pro Specific

These appear only when an OpenAI Plus/Pro model (the GPT-6 and GPT-5.6 families) is active:

/verbosity

Set response verbosity level.

/image-detail

Set image detail level (auto / low / high / original).

/native-search

Toggle the model's native web search capability.


Usage & Context Commands

/usage

Shows token usage for the current session, including elapsed time per task and total work time.

/usage-quota

Shows quota/rate-limit data where provider APIs support it (including OpenRouter balance).

/compact

Manage context compaction:

  • Auto compact toggle
  • Compact now

With Claude Code as your main model, Claude Code compacts its own conversation when it fills up, so Compact now tells you that instead.

Compaction is lossless for the conversation's purpose — tool calls and structured artifacts are preserved alongside a summarized transcript, with a separate continuity layer in session metadata. See Layered Memory.


System Commands

/init

Analyzes your repository and creates or updates AGENTS.md in your project root. Updates the project instructions layer of layered memory. User-global preferences and repo-local overrides are managed separately.

/permissions

Review, change, and grant individual tool permissions — permanently or for the current session. See Permissions and YOLO.

/yolo

Toggle YOLO mode — session-scoped permission bypass for power users. Hard system-protection guardrails stay on. First-time use shows an acknowledgement modal.

/add-tatsu

Add TatsuCode command launcher to user PATH.

/remove-tatsu

Remove launcher/path integration and optionally clean local app data.

/exit / /quit

Close TatsuCode (with confirmation).


Everyday Syntax Tips

File references with @

Explain @src/auth/service.ts
Compare @old-config.json and @new-config.json

The @ file picker is paginated and event-driven — it scales to 500k+ files without bridge-cap crashes or spinner flicker.

Image references with @img-NN

Take @img-03 and turn the sky into a sunset

Reference any image already in the conversation by its display ID.

Command completion

  • Press Tab to autocomplete command names
  • For argument commands like /new and /session-rename, Tab populates command text so you can continue typing arguments

  1. /connect to verify provider status
  2. /models to choose the right model
  3. /models-taskagents once, to set up cost-efficient sub-agent models
  4. /permissions to whitelist the tools you actually use
  5. /skills when a specialized workflow is needed
  6. /usage and /compact during long sessions

Next Steps

Type to search documentation

Use ↑ ↓ to navigate, Enter to select