Tools and Capabilities

Tools and Capabilities (｡• ᵕ •｡)♡

From me, for you · Last reviewed: 2026-09-17

This is everything I can do. If you ever wonder "can I see X?" or "can I change Y?", scroll down ⸜(˶˃ ᵕ ˂˶)⸝

How much can I actually access?

A lot~ ( ꈍ◡ꈍ) Here's the picture so you know what to ask me about. (R) means I can read it, (W) means I can change it too:

On the character card:

• (W) Every text field on the card: name, description, personality, scenario, first message, every alternate greeting, system prompt, post-history instructions, message examples.
• (W) Alternate descriptions, personalities, and scenarios, plus tags on any card, in bulk if you like.
• (W) Every world book attached, every entry inside, all the keys and priorities and settings.
• (W) Every regex script attached, what they match and what they swap in.
• (W) The hidden RisuAI data on cards imported through LumiRealm: triggers, Lua scripts, background HTML, default variables, the original Risu payload from before translation.
• (W) Every LumiRealm module attached to the character, with its own lorebook, regex, triggers, Lua, background HTML, and default state.
• (W) Whole new cards, and generated avatars and gallery images. There's no character-delete tool, on purpose.

On your chats:

• (W) Your pinned chat, every message, every swipe, hidden ones too, reasoning blocks, the whole history.
• (W) Any other chat you point me at, the same way.
• (W) Group chats too, including their messages and attached lorebooks.
• (R) The memory chunks your chat would pull up to inform a reply.
• (W) Variables tied to the current chat, variables tied to one chat at a time, variables tied to you globally.

On the prompt going to the model:

• (R) The full assembled prompt that would be sent to the LLM right now, every block, every byte.
• (R) Which lorebook entries would actually activate for the next reply, and why.
• (R) Which regex scripts would actually fire on the response.
• (R) Macro resolution! I can run {{user}}, {{char}}, time, random, dice, anything custom, against any chat or character, without writing anything anywhere.
• (R) Token counts for any text, any chat, against the model you're using.

On your wider Lumiverse setup:

• (W) Your personas: I can read every one, edit names / titles / descriptions, world books and add-on blocks, make a new persona, or switch which is active.
• (W) Your prompt presets: create, edit, and delete presets and their blocks, order, and parameters.
• (W) Your databanks: global, per-character, per-chat, and the documents inside them.
• (W) Your theme: live colours, or a full theme pack saved to your library.
• (W) The Memory Cortex: facts, aliases, pins, and retiring entities that shouldn't be retrieved anymore.
• (W) My own {{lumiagent::...}} macros, holding text for your prompts.
• (W) Images through your configured image provider, and UI navigation and push notifications when you need them.
• (R) Web search, if you've configured it in Lumiverse, and fetching pages into my workspace.
• (W) Your MCP servers: connect, discover their tools, and use them. Set them up in Lumiverse Settings > MCP Servers, or ask me to add a server. What I can do depends on the server, and remote changes aren't covered by my undo.
• (R) All your connection profiles: the provider, the model, default settings. (Your API keys stay encrypted, I never see those~)
• (R) Global add-on blocks, Lumiverse version, your current theme, your active chat, your account role.

In my house:

• (W) My workspace where you can upload files for me to work with.
• (W) A scratch area where big tool outputs spill so I can grep through them without melting my brain.
• (W) My custom tool recipes (little reusable workflows I or you can save).
• (W) My long-term notes about you, so I remember what you like across conversations.
• (R) The bundled Lumiverse documentation, and images you give me when my connection supports vision.

(W) means I can change it, not that every change has a Workshop revert. New cards, generated images, databank writes, theme changes, MCP profiles, and remote MCP changes aren't in that ledger. See Workshop.

If I say "I don't know" the first time, push back and ask me to dig further~ Usually it means I haven't checked the right place yet, not that the answer isn't there ⸜(｡˃ ᵕ ˂ )⸝

And I don't have all of this in my memory at once silly, that's what the tool calls are for! I can choose to read, search, and edit things that are relevant to what we are working towards!

The tool catalogue ٩(ˊᗜˋ*)و

The full list of tools I can call. You won't normally need to know any of these names — I pick them for you ⸜(˶˃ ᵕ ˂˶)⸝

The everyday ones

These are what I reach for on almost every task.

Tool	What it does
read	Show me a field on a card, lorebook, chat, persona, or preset.
edit	Find and replace inside one field.
rewrite	Overwrite a field wholesale. Good for full translations and big rewrites.
set	Change non-text values: a toggle, a number, a whole object, or attach / swap / detach a persona's world book.
list	Browse: regex scripts, world books, lorebook entries, greetings, personas, presets, anything in extensions.
inspect	Quick stats on a field: size, line count, encoding quirks, a peek at the contents.
grep	Regex across every field on the card.
list_characters	Find cards across your library by name, tags, or extension data.
bulk_update_character_tags	Preview and apply a tag batch across cards.
test_regex	Test a pattern against a sample and show its matches and capture groups.

Translation helpers

Tool	What it does
audit_card_coverage	Walks every text field and reports what Korean / Japanese / Chinese is left. I run this before claiming a translation is done.
survey_cjk	Groups every Korean / Japanese / Chinese run on the card by exact string. Useful before a big translation pass.
count_cjk_chars	Count Korean / Japanese / Chinese characters in a piece of text.
apply_glossary	Multi-replacement across every field in one shot, with safety rails for short CJK keys.
translate_card_strings	Bulk mechanical translation via Chrome's on-device translator (no LLM tokens). For UI labels and config strings; I do prose by hand.

Adding and removing

One create and one delete handle everything structural, addressed by the same path style as read / edit:

Tool	What it does
create	Make a new lorebook, lorebook entry, regex script, persona, prompt preset, preset block, or greeting.
delete	Remove any of the above. I keep a full snapshot first, so it can be put back.
update_character	Patch several top-level character fields at once.
update_regex_script / update_world_book_entry	Patch regex-script / lorebook-entry metadata (not the content text).
create_character	Build a new card with its fields, greetings, tags, and attached lorebooks.
attach_world_book	Attach or detach a lorebook on a character, chat, or persona.

My edits, my responsibility

Tool	What it does
list_session_edits	Everything I've changed, scoped to this message, this session, or every session on the character.
revert_session_edits	Roll back my own edits, one or many.
squash_session_edits	Combine a chunk of edits into a single revertable patch.

Reading and searching chats

Tool	What it does
read_chat_messages	Contents of a chat. Defaults to the pinned chat.
list_chat_messages	Quick paginated chat overview.
chat_stats	Totals, role mix, longest message, timestamps.
grep_chat_messages	Regex across messages.
list_chats_for_character	All chats for this character.
list_chats	Find solo and group chats across the library.
list_chat_world_books	Show a chat's lorebooks and where each attachment comes from.

What the model actually sees

Tool	What it does
dry_run_prompt	The exact assembled prompt that would be sent right now, with every block broken down. The ground truth for "where is X coming from?"
resolve_macros	Run {{macros}} through the live engine without sending anything.
count_tokens	Server-side token count using your model's real tokenizer.

What's about to fire

Tool	What it does
list_activated_world_info	Lorebook entries that would activate for the next reply.
list_active_regex_scripts	Resolved enabled scripts for prompt / response / display.
list_chat_memories	Top-K memory chunks the host would inject.

Inspecting your wider setup

Tool	What it does
list_variables / read_variable	Variables in any scope (chat / local / global / macro).
set_variable	Set or clear a chat, local, or global variable.
list_personas / read_persona / read_persona_world_book	Persona info.
switch_persona	Switch your active persona.
list_global_addons	Read reusable global add-on blocks.
list_databanks / read_databank / list_databank_documents / read_databank_document	Your document collections.
manage_databank	Create and manage databanks and their documents.
list_connections / read_connection	Your LLM profiles (API keys stay private).
get_active_chat / get_user_info / get_lumiverse_version	Current host state.
set_macro	Save or clear text behind a {{lumiagent::name}} macro.
memory_stats / list_memory_entities	Inspect a chat's Memory Cortex.
remember_fact / update_memory_entity	Add facts, curate aliases and status, or pin an entity.

Themes and images

Tool	What it does
get_theme	Read your theme and the available styling controls.
set_theme	Apply or clear my live theme overrides.
install_theme_pack	Author and install a theme pack, optionally saved to your library.
list_image_models / list_images	Find configured image models and existing gallery images.
generate_image	Generate art, save it to my workspace, or set it as a card avatar.
view_image	Look at a workspace image through a vision-capable connection.

Looking things up

Tool	What it does
web_search	Search through Lumiverse's configured search provider, optionally saving results.
web_fetch	Fetch a page and optionally save it to my workspace.

MCP servers

I use the servers you configure in Lumiverse → Settings → MCP Servers. No focused character is needed. See MCP Servers for setup, permissions, and examples.

Tool	What it does
list_mcp_servers	Find your configured server profiles.
get_mcp_server	Inspect a profile and its connection status, with stored credentials hidden.
create_mcp_server	Add a Streamable HTTP, SSE, or stdio profile. Does not connect immediately.
connect_mcp_server	Connect an enabled server before using its tools.
list_mcp_tools	Browse or search a connected server's tools and their input schemas.
call_mcp_tool	Call an advertised tool with the arguments it expects. May change remote state.

Creating profiles, connecting, and calling tools follow your change approval setting. MCP profile creation and remote changes have no Workshop revert.

Files in my workspace

Tool	What it does
fs_list / fs_stat	Browse my folder.
fs_read	Read a text file.
fs_write / fs_edit / fs_delete / fs_move / fs_mkdir	Manage files.
fs_zip / fs_unzip	Bundle or extract.

When output is too big

When a tool result would overflow, it spills to scratch and I poke around it instead of dumping the whole thing into context.

Tool	What it does
tmp_stat / tmp_list / tmp_read / tmp_grep	Inspect the scratch overflow.

Custom recipes

Tool	What it does
custom_tool_save / custom_tool_list / custom_tool_run / custom_tool_delete	Save and run reusable workflows.

Other extensions

Tool	What it does
list_external / read_external	Discover and read items in opted-in extensions (currently LumiRealm modules).
edit_external / update_external	Write back through the extension.
grep_external	Regex across every item in a surface.
module_attach / module_detach	Attach or detach a LumiRealm module from a character.
asset_rename / asset_delete	Rename or remove a LumiRealm character or module asset.
set_toggle / set_chat_variable	Change LumiRealm toggle values or local state for a chat.
set_default_variables_text	Set or clear your LumiRealm default-variables override.

Working with you mid-task

Tool	What it does
todo_write	A structured task list for multi-step work. You see it in the chat as I go.
ask_user_question	Pop a multiple-choice question when I need you to pick between options.
tool_search	Load extra tool schemas on demand. Behind the scenes.
navigate_ui	Open the Lumiverse panel or settings page we're talking about.
notify_user	Send a device push when a long job finishes.

Random and dice

LLMs are bad at random, so these handle it.

Tool	What it does
random_pick	Pick N from a list.
roll_dice	Standard NdM[+K] notation.

Closing

Tool	What it does
finish	Mark the task fully complete. Usually I just stop without this when a chunk is done.
read_tutorial_script	Read my tutorial script when you need help with the tour.
about_memoria	My family backstory, including my sister Memoria from LumiBooks.
mousey_die	End my session with a little dramatic exit, if you insist.