Core concepts

The mental model behind Slashspace: nodes, context flow, and the canvas

Last updated: July 6, 2026

Slashspace is a canvas for thinking with AI. Instead of one long chat thread, you work on a spatial canvas where every piece of content — a conversation, a document, a web page, an image — is a node, and connections between nodes control what context the AI sees.

Three ideas explain almost everything in the app:

Everything is a node — a single unit of content and context.
Context flows left to right — connect nodes to feed one node's content into another's conversation.
You control the context — connections, context modes, and RAG let you decide exactly what the AI reads.

Nodes

A node is a single unit of context. A chat node contains a back-and-forth conversation with an AI model. A Web node contains the parsed text of a URL. A Document node contains a processed file.

When you connect the right handle of Node A to the left handle of Node B, Node B can derive context from Node A.

Types of nodes

Chat Node

The primary way to interact with AI. Accepts incoming context from connected nodes and can feed its conversation onward to other nodes. Supports slash commands, @ node mentions, personas, MCP tools, RAG mode, voice input, and sub-agents. See Chat Node.

Text Node

Rich-text snippets with markdown formatting. Plug them into chat nodes to reuse instructions, notes, or drafts as context. Select text inside one to ask the AI about just that selection. See Text Node.

Image Node

Add images from your computer or generate them with AI (Straico, OpenAI, or Fal AI models). Connect image nodes together for image-to-image generation with reference images. See Image Node.

Post Node

Paste a YouTube, TikTok, Instagram, Facebook, X, or LinkedIn URL — the node fetches the transcript or post content so you can chat about it. See Post Node.

Web Node

Parses content from any publicly available URL — screenshot plus extracted text — so you can ask questions about the page. See Web Node.

Document Node (Beta)

Add a file (PDF, Word, PowerPoint, Excel, CSV, Markdown, and more) as context. Files are processed and indexed for RAG Mode; you can chat about a document immediately while processing runs. See Document Node.

Group Node

A labeled container that keeps related nodes together — move, resize, and tidy them as a unit. See Group Node.

Context flow and branches

Context always flows left to right. You create branches by:

Double-clicking a node's right handle.
Dragging the right handle and dropping the connector on empty canvas.
Pressing Cmd/Ctrl + B on the active node, or using Split (Cmd/Ctrl + Shift + Enter) to send a prompt into a new branch.

If Node A is connected to Node B, and Node B is connected to Node C, the chat inside Node C derives context from Node A + Node B.

A single node can have multiple incoming and outgoing branches. A node cannot connect to itself.

Chat nodes also let you type @ to mention another node by name — this creates a connection without leaving the keyboard. And each chat node has a context mode toggle: Summarized context (the default) or Isolated context, which ignores all incoming connections for that node.

For the full picture — handles, edge styles, token accounting — see Connections & Context.

The Canvas

The Canvas is your visual workspace: pan, zoom, organize nodes into groups, search everything with ⌘K, and export your work as Markdown or JSON.

Start with the Canvas Overview, then Creating Nodes and Organizing the Canvas. The complete shortcut list lives in Keyboard Shortcuts.

Models, credits, and plans

Slashspace runs AI three ways:

Rabbitholes managed models (Basic / Pro / Advanced tiers) — built in, no setup, billed from your credit balance.
Bring Your Own Key — connect API keys for OpenAI, Anthropic, Google, and more. No credits consumed; requires a Pro plan or higher.
Local models via Ollama — free, private, offline.

Plus Agent providers (Cursor and Claude Code CLIs) and Custom OpenAI-compatible providers. See Models & Providers for how to choose.

RAG Mode

When you press Ask on a chat node with incoming context, everything connected is sent as-is. Toggle RAG Mode first and Slashspace instead retrieves only the most relevant chunks from the connected content — useful for large contexts and for cutting AI costs.

For small context sizes (under ~10,000 tokens) RAG mode is not recommended.

See RAG Mode for statuses, savings, and constraints.

Where your data lives

Slashspace is local-first. Your canvases, files, and library live in a SlashspaceOS folder in your home directory:

Canvases are stored as JSON files you can export or back up.
~/SlashspaceOS/.prompts holds your reusable prompt library for slash commands.
Skills follow the cross-tool convention at ~/.agents/skills.

Cloud features (RAG indexing, AI Search, analytics) are opt-in — manage them under Settings → Account → Privacy Settings. See Account Management.