Local CLI + MCP server
Give your AI access to Space.
Work from the terminal—or connect Codex and Claude directly through MCP. Search your decks, create cards, and improve what you already know without leaving the conversation.
Install the Space app on Mac, Windows or Linux first. The CLI reads the database it keeps on your machine. No login, no API keys.
$ space deck list
┌──────┬─────────────────────────────┬───────┬──────┐
│ ID │ Name │ Cards │ Due │
├──────┼─────────────────────────────┼───────┼──────┤
│ ck3u │ Japanese Kanji — N5 │ 104 │ 12 │
│ f02a │ FSRS Paper Notes │ 37 │ 3 │
│ 9a1e │ Rust Ownership & Lifetimes │ 58 │ 0 │
└──────┴─────────────────────────────┴───────┴──────┘ Install
New here? Install the Space app first and open it once, then pick one below.
Homebrew
macOS & Linux
brew install space-org/tap/space-cli Recommended — updates automatically with `brew upgrade`.
★ Recommended for youcurl installer
macOS & Linux
curl -fsSL https://raw.githubusercontent.com/space-org/space-cli/main/scripts/install.sh | sh Drops the binary into ~/.local/bin. Works on any shell.
★ Recommended for youManual download
Windows, macOS, Linux
Open latest release →Grab the archive for your platform and put the binary on your PATH.
★ Recommended for youAsk. Space does the rest.
space mcp gives supported AI clients a small,
purpose-built set of tools for your local flashcard library.
No exports to shuffle around. No Space API key.
The server runs locally over stdio and makes no network calls. Your AI client still applies its own approval and privacy settings.
list_decks get_deck search_cards create_deck create_cards update_card One-time setup
Pick your client.
$ codex mcp add space -- space mcp $ claude mcp add --scope user space -- space mcp Claude Desktop
{
"mcpServers": {
"space": {
"command": "space",
"args": ["mcp"]
}
}
} Start a new client session after setup, then ask it to list your Space decks. Approve write tools only when you want changes made.
Codex, Claude Code, and Claude Desktop can launch this stdio server. ChatGPT web and Claude web cannot connect to it directly.
Quickstart
space deck list Show every deck with card and due counts space deck stats ck3u Retention, maturity mix, due today space card search "past tense" Full-text search across all cards space card show 7f2a Render a single card in the terminal space deck export ck3u --format csv Export a deck to CSV
Run space --help for the full
reference.
deck,
card and
group each support
create,
show,
list,
edit,
delete,
search,
stats and
export.
Pair it with AI.
Pipe structured data into any LLM. Get back analyses, explanations, new cards. A GUI cannot touch this.
Find confusing word pairs
Export a deck and let the AI scan every card at once. It spots false friends and overlapping meanings you miss one card at a time.
$ space deck export ck3u --format json | claude \
"Which of these Spanish words have similar meanings
I could easily confuse? Generate mnemonics for the
most confusing pairs." Explain a tough card
When a card refuses to stick, ask for a deeper explanation. Linking new knowledge to old builds more retrieval paths.
$ space card show 7f2a | claude \
"Explain this concept with a hands-on example and
an analogy to something I already know as a developer." Derive new cards from existing data
Answering before seeing the solution anchors knowledge deeper. Follow-up questions hit that mechanism.
$ space deck export f02a --format json | claude \
"Analyze these system-design flashcards and create 10
follow-up questions. Format: JSON array with 'front'
and 'back' per card." Any AI works. Claude, ChatGPT, Ollama, LM Studio: the CLI hands over the data, you pick the tool. With a local model, nothing leaves your machine.
Sync across every device.
Cards you create in the CLI show up on your phone, tablet and desktop. Cards you learn on the go are queryable from the CLI right away.
Changes queue locally first. Next time the Space app is online, they sync to every device. No manual sync, no import or export, no cloud login in the CLI.
Open source
Issues, contributions and a full command reference live on GitHub.