Core concepts

A chat app keeps everything in one scrolling thread, and the model reads whatever happens to be in it. Slashspace replaces the thread with a canvas. Your conversations, notes and sources sit side by side as nodes, and you draw connections to decide exactly what each conversation reads.

Once the ideas below click, every other page is detail.

Canvas

A canvas is an infinite surface you pan and zoom around. Give each project or line of research its own canvas, so its sources and conversations stay together, and you come back weeks later to a map instead of a transcript.

You can keep many canvases. They live in the sidebar, where you can sort them into folders (the sidebar calls canvases spaces). See Canvases and folders.

Nodes

A node is one unit of content and context: a conversation, a note, a web page, a video, a document, an image or a sketch. Anything you put on the canvas as a node is something a chat can read.

Node	What it holds
Chat node	A conversation with an AI model. It reads whatever you connect to it.
Text node	Your own notes, instructions or drafts, in rich text.
Web node	The content of a web page, captured from its URL.
Post node	A YouTube, TikTok, Instagram, Facebook, X or LinkedIn post, with its transcript when one is available.
Document node	A PDF, Word, PowerPoint or Excel file, or a text or code file.
Image node	An image from your computer, or one generated in Image mode.
Scratch pad	A freehand sketch. Connected chats see it as an image.
Group	A labeled container for related nodes. Connect it to a chat to pass along everything inside.

Connections are context

A connection decides what a chat reads. Every node has a handle on its right edge, and chat nodes also have one on their left. The right handle gives and the left handle takes: drag from a node's right handle to a chat's left handle, and the chat reads that node's content with every message you send.

So context flows left to right: sources on the left, the conversations that use them to their right.

Context also flows through chains. If A connects to B and B connects to C, the chat in C sees both A and B, not only its direct neighbor. A node that reaches a chat by more than one path counts once.

You don't have to drag. Type @ in a chat's message box and pick a node: the mention adds a connection from that node to the chat, and the connection stays after you send. See Connections and context.

Context modes

Every chat node has a context mode, which sets what it sees beyond its own conversation.

• Summarized context (the default). A chat with no connections gets a short summary of every other node on the canvas, so it knows what is there. In Agent mode, the AI can then open the full content of a node that looks relevant. Once you connect something, the chat reads only what is connected.
• Isolated context. The chat ignores the rest of the canvas, including any connections, and works only from its own conversation. Use it when you want a clean-slate answer on a busy canvas.

You switch modes from the chat's left handle while it has no connections. See Connections and context.

Branches

A long conversation drifts. You ask a side question, the answer takes over, and the thread you cared about gets buried. On a canvas you fork instead: click Submit in new branch, or press ⌘⇧Enter (Ctrl+Shift+Enter on Windows and Linux), and your message starts a new chat to the right, connected to the one you were in.

Because a branch is connected to the chat it came from, it sees that conversation and everything feeding it, while the original thread stays untouched. See Branching.

Chat modes

Each chat node runs in one of four modes, picked from the mode menu in its message box. The mode decides what the AI can do.

Mode	What it does
Agent	The default. The AI can use tools: your connectors, sub-agents that work in their own chat nodes, and reading other nodes on the canvas.
Chat	A plain conversation with no tools.
Image	Generates images from your prompt. Connected images can serve as references, and each result lands in its own Image node.
Orchestrator	Builds on the canvas for you. Give it a goal and it adds chat, text, web and post nodes, connects them, and edits their titles and text.

Where the AI runs

Each chat node has its own model, chosen from the model menu in its message box, so two chats side by side can put the same question to different models. Out of the box, chats use Auto, which picks a Slashspace hosted model for each message. Slashspace's hosted models and hosted tools, such as image generation, are paid from your hosted wallet, and your plan includes some wallet balance.

You can also use your Claude or ChatGPT subscription (through Claude Code or Codex), your own API keys from providers such as OpenAI, Anthropic and Google, local models through Ollama, or any OpenAI-compatible endpoint. None of these draw from the wallet. See Choosing a model.

Where your data lives

Slashspace is local-first. Each canvas, with its nodes, conversations and connections, is saved as a folder inside SlashspaceOS in your home directory, and your sidebar folders are real folders there. Content leaves your computer when a chat sends it to a cloud model, or when you use a hosted feature such as capturing a web page. See Privacy and your data for exactly what is sent and when.

Next

• Adding nodes: every way to put content on the canvas.
• Connections and context: handles, chains, groups and context modes in detail.
• Chat node: the message box, modes, models and tools.