Can I use it without an account?

Yes. You can send messages, use the preloaded task cards, and see generated UIs without signing in. The number of messages per day is capped (10 by default) and nothing is kept in history. Keeping history, creating or editing tasks, saving memos, and creating share links need a free account.

What does a task button actually do?

Pressing a card sends a message shaped like “【タスク】Summarize”, carrying whatever you typed in the box as the situation section. Behind it, the task's prompt template, response rules, output template, and input and output examples go to the model, and you can read all of them under the “Prompt” disclosure after sending.

When does a generated UI appear? Can it appear uninvited?

Only when you explicitly ask for something visible, such as a chart, a diagram, or an interactive demo. Comparisons, procedures, calculations, and code examples come back as text as usual, and writing “text only” or “no diagram” keeps it that way.

What can 3D do, and what can it not do?

Asking for 3D, a solid shape, or Three.js returns a Three.js screen with a scene, camera, light, and geometry that you can drag to orbit. Three.js is the only library available: no external CDNs, and no textures or 3D models loaded from a URL. The frame is between 160 and 900 pixels tall.

Can other people see my conversations?

Not unless you create a share link yourself. Once you do, anyone with the URL can read that conversation and press “Continue this chat” to keep going in their own copy. Temporary chats are never stored and cannot be shared at all.

CHATCORE-AI / AI CHAT

Repeated instructions become buttons. Answers can be things you operate.

ChatCore-AI keeps the instructions you send again and again as task cards you launch with one press. Ask for a chart, a diagram, or a 3D model and the answer comes back as a working UI rendered inside a sandbox instead of a wall of text. You can try it without signing in.

Open the chat See what it does

Works without an account
Four models to switch between
14 task cards from the start

WHAT YOU CAN DO

Launch it, render it, keep it.

Task buttons

Keep the instructions you send often as cards you launch with one press. Fourteen come preloaded, and once you sign in you can add and rewrite your own.

Pressing a card sends it as “【タスク】Summarize”, and whatever you typed in the box travels with it as the situation section
The task body (prompt template, response rules, output template, input and output examples) stays readable under the “Prompt” disclosure after sending
The + button creates a new task: write the title and body yourself, or let AI assistance draft it
Pencil to edit, trash to delete, and hold a card to pick it up and reorder it (signed in)
Pressing “Use in chat” on a shared prompt drops that public prompt straight into your task list

Generated UI and 3D

Only when you ask for a chart or a diagram does the answer come back as a working UI. Ask for 3D and it builds a Three.js scene instead.

2D results are charts, tables, timelines, and small demos you can interact with in place
3D results build a scene, camera, light, and geometry, and you can drag to orbit the view
At most three UIs per message, each in a frame between 160 and 900 pixels tall
If the result comes back incomplete, one automatic repair pass runs before you see it
Write “text only” or “no diagram” and it answers in prose instead

Around the conversation

The controls for keeping, passing on, and redoing an answer are all in the same place.

Switch between four models: fast, balanced, deep thinking, and careful writing
Web search runs only when the AI judges that fresh information is needed, and sources appear as chips in the answer
Save an answer as a memo or copy it; deliverables such as an email body come back in their own copy block
“Share chat” issues a link, and the person who opens it can press “Continue this chat” to keep going in their own copy
Editing a message or regenerating an answer adds a version you can move through with ‹ 1/2 ›
Temporary chat keeps nothing in history, and a project applies its custom instructions to every chat inside it

WHAT IT COVERS

How you ask decides what comes back.

SHAPE OF THE ANSWER

Text (default)Explanations, comparisons, procedures, calculations, and code come back as prose and Markdown.
2DAsking for a visualization, diagram, chart, flow, timeline, or interactive demo produces a UI you can use in place.
3DAsking for 3D, a solid shape, Three.js, or rotation produces a Three.js scene.

WHAT YOU CAN SEND

The message boxUp to 30,000 characters per message.
FilesAttach PDF, Word, Excel, PowerPoint, text, and code files: five at a time, 1MB each.
Lookup togglesThe add menu turns on lookups against your memos and personal context, or against shared prompts (memos require signing in).

Models you can switch between

GPT-OSS 120B (fast responses)
GPT-5.6 Luna (balanced)
Qwen 3.6 27B (deep thinking)
Claude Haiku 4.5 (careful writing)

Before you start, know this

A generated UI is display-only. It runs inside a sandboxed iframe with no network access, no storage, and no access to the page around it; images and fonts must be embedded.
Three.js is the only library available for 3D, and it is served by ChatCore-AI itself. Libraries from external CDNs cannot be loaded.
A generated UI appears only when you ask for one. Comparisons, procedures, calculations, and code examples come back as text as usual.
The AI decides whether to run a web search. There is no manual switch that forces one.
Attachments are documents and code (up to five files at a time, 1MB each). Images cannot be attached.
You can send messages without signing in, but the number of messages per day is capped (10 by default) and the conversation is not kept in history.
A temporary chat is never stored and cannot be shared. A share link you do create can be opened by anyone who knows the URL.

HOW IT WORKS

Type it, press it, keep the result.

Write what you need
Type your situation into the box on the home page. No account is needed this far, and you can attach document or code files if they help.
Press a task, or just send it
If a task card fits, press it: the card's instructions and whatever you typed are sent together. If none fits, send the message as it is.
Keep what came back
Save an answer as a memo, share it as a link, or regenerate it to compare alternatives. Saving and sharing need a free account.

How to ask

When you want a working UI: Say “as a chart”, “as a diagram”, or “make it interactive”. Without asking, you get prose.
When you want 3D: Say “in 3D”, “as a solid”, or “with Three.js”. You can drag to orbit whatever it builds.
When you want text only: Add “text only” or “no diagram” and it answers in prose without building a UI.

Price and prerequisites

Price: Free. No credit card is involved.
Without an account: Send messages, use the preloaded task cards, and view generated UIs. The daily message count is capped and nothing is kept in history.
With a free account: Keep your history, create and reorder your own tasks, save memos, create share links, and group chats into projects. Sign up with an email address or with Google.

FREQUENTLY ASKED QUESTIONS

Questions people ask before the first message.

Can I use it without an account?: Yes. You can send messages, use the preloaded task cards, and see generated UIs without signing in. The number of messages per day is capped (10 by default) and nothing is kept in history. Keeping history, creating or editing tasks, saving memos, and creating share links need a free account.
What does a task button actually do?: Pressing a card sends a message shaped like “【タスク】Summarize”, carrying whatever you typed in the box as the situation section. Behind it, the task's prompt template, response rules, output template, and input and output examples go to the model, and you can read all of them under the “Prompt” disclosure after sending.
When does a generated UI appear? Can it appear uninvited?: Only when you explicitly ask for something visible, such as a chart, a diagram, or an interactive demo. Comparisons, procedures, calculations, and code examples come back as text as usual, and writing “text only” or “no diagram” keeps it that way.
What can 3D do, and what can it not do?: Asking for 3D, a solid shape, or Three.js returns a Three.js screen with a scene, camera, light, and geometry that you can drag to orbit. Three.js is the only library available: no external CDNs, and no textures or 3D models loaded from a URL. The frame is between 160 and 900 pixels tall.
Can other people see my conversations?: Not unless you create a share link yourself. Once you do, anyone with the URL can read that conversation and press “Continue this chat” to keep going in their own copy. Temporary chats are never stored and cannot be shared at all.

Send one message and see what comes back.

Sending works without an account, though the daily count is capped and nothing is kept. Sign up when you want history, your own tasks, memos, and share links.

Open the chat