Daiso CLI

search daiso, convenience stores (GS25, Seven-Eleven, CU, Emart24), marts (Lotte Mart), retail (Olive Young), and cinemas (Megabox, Lotte Cinema, CGV) via the daiso CLI or MCP endpoint. use the CLI for direct execution when npx is available. use the MCP endpoint when configuring an AI app or when the host environment doesn't support shell commands.

intent

this skill queries product inventory, store locations, cinema showtimes, and seat availability across Korean retail, convenience, and cinema chains. run this when the user asks for a product at a specific location, store finder results, movie times, or seat maps. prefer CLI execution for speed and direct output. switch to MCP only when shell access is unavailable or the host app explicitly requires remote MCP.

inputs

required

• service: target chain (daiso, gs25, seveneleven, cu, emart24, lottemart, oliveyoung, megabox, lottecinema, cgv)
• query: korean product name, store keyword, or theater name
• npx or node: CLI requires node.js and npm installed (check which npx)

optional

• --json: return structured JSON instead of human-readable text
• --limit N or --pageSize N: cap result count (default varies by endpoint)
• --keyword: product search term
• --storeKeyword: filter results by store location or name
• --playDate YYYYMMDD: cinema showtimes (compute today in KST if omitted)
• --theaterCode: CGV or Lotte Cinema code for direct theater lookup
• --area: region filter (e.g., 서울, 부산)
• --storeName: specific store (e.g., 강변점)
• product_id: Daiso product ID for inventory drill-down

external connections

• daiso npm package: npx daiso CLI wrapper (auto-installed via npx)
• MCP endpoint: https://mcp.aka.page (remote MCP server, used when host app supports it)
• live retail APIs: Daiso, GS25, Seven-Eleven, CU, Emart24, Lotte Mart, Olive Young, Megabox, Lotte Cinema, CGV (subject to rate limits, auth tokens, IP blocks)

environment assumptions

• user can run shell commands (for CLI path) or has an app configured for MCP
• korean keyboard input or copy-paste of korean text supported
• network access to public daiso endpoints and cinema APIs
• KST timezone for cinema date computation

procedure

1. identify the service and query intent

   • input: user request in korean or english
   • extract service name (daiso, gs25, seveneleven, emart24, etc.)
   • extract search term (product name, store location, movie title, theater name)
   • output: service slug, query string, optional filters

2. choose execution path

   • input: availability of npx, host app MCP support, user request
   • if npx available and user not configuring an app, proceed to step 3 (CLI)
   • if host app requires MCP or npx unavailable, jump to step 8 (MCP)
   • output: chosen path (CLI or MCP)

3. run health check (CLI path)

   • input: none
   • execute: npx daiso health
   • output: daiso service status
   • if health check fails, report failure and do not proceed; suggest checking network and retrying

4. construct the CLI command

   • input: service, query, optional filters, user's need for JSON
   • match against common commands in section "common commands" or references/cli-command-map.md
   • if exact command exists, use it; if not, fall back to npx daiso get /api/<service>/<endpoint>
   • add --json if user is comparing results, integrating with another tool, or needs machine-readable output
   • output: full command string ready for shell

5. execute the command

   • input: command string from step 4
   • run in shell (bash, zsh, or equivalent)
   • capture stdout and stderr
   • output: raw JSON or formatted text response

6. parse and summarize results

   • input: raw response from step 5
   • if --json was used, parse JSON and extract user-relevant fields: product name, price, store name, address, stock count, movie title, showtime, seat availability
   • preserve product IDs, store codes, theater codes, movie codes for potential follow-up queries
   • note if data source is live external service (subject to change)
   • output: human-readable summary or structured JSON blob

7. report outcome and next steps

   • input: summary from step 6
   • if results found, display them; if no results, suggest broader keywords or nearby areas
   • if user may need follow-up (e.g., seat booking, inventory check at specific store), preserve IDs and offer next command
   • output: formatted response to user

8. use MCP endpoint (fallback or explicit request)

   • input: service, query, filters (from step 1)
   • provide user with MCP URL: https://mcp.aka.page
   • construct equivalent GET /api/<service>/<endpoint>?param=value path
   • output: MCP endpoint details and example curl or API call
   • do not execute MCP call; user's host app or manual curl will do so

decision points

if npx and node.js are available

• then: run CLI commands directly (steps 3-7)
• else: use MCP endpoint (step 8) and inform user of environment limitation

if user request includes both product and location (e.g., "콜라 강남역")

• then: prefer inventory lookup over product-only search (e.g., gs25-inventory 콜라 --storeKeyword 강남역)
• else: run product search first, then offer inventory drill-down with location

if user asks for cinema showtimes but omits date

• then: compute today's date in KST as YYYYMMDD and use as --playDate
• else: use provided date (do not copy example dates from docs)

if user provides area/region but no specific store

• then: search products in that region; ask for narrower store name before inventory lookup
• else: run query as-is and report results by store location

if CLI command fails (no results, service down, network error)

• then: run npx daiso health to check service status
• if health passes, retry with narrower query or different keywords
• if health fails, report the failing service and suggest MCP as fallback
• else: do not invent results; report the exact command and error

if user is configuring an AI app or explicitly requests MCP

• then: skip CLI path and jump directly to MCP details (step 8)
• else: use CLI

if Seven-Eleven inventory is requested

• then: use raw GET fallback npx daiso get /api/seveneleven/inventory --keyword <product> --storeKeyword <location>
• else: use named command if available

if user asks for Daiso inventory but provides only product name (no store/area)

• then: prompt user for area or store keyword before querying inventory
• else: run inventory with provided keyword filters

output contract

CLI JSON response (--json)

• format: valid JSON object or array
• schema: varies by service and endpoint; always includes id, name, price (if product), address (if store), showtime (if cinema), stock (if inventory)
• example daiso product: {"id": "1034604", "name": "수납박스", "price": 5000, "category": "storage"}
• example store: {"code": "gs25_001", "name": "강남역점", "address": "서울 강남구...", "distance": "0.2km"}
• example cinema: {"code": "cgv_001", "name": "서울강남", "showtimes": [{"movieId": "...", "time": "14:00", "seats": 45}]}

CLI text response (no --json)

• format: human-readable list or table
• content: product names, prices, store names, addresses, stock counts, showtimes, seat info as applicable
• line breaks and spacing for readability

MCP endpoint reference (step 8)

• format: plain text or markdown
• content: https://mcp.aka.page and example GET path
• example: GET /api/daiso/products?query=수납박스&limit=5

error response

• format: error message with command used
• content: "no results for in ; try broader keyword" or "service currently unavailable; check health with npx daiso health"
• always preserve the command so user can retry or adjust

file location

• stdout: all CLI output (no file creation by default)
• if user pipes to file (e.g., npx daiso products ... --json > results.json), that's user's responsibility

outcome signal

• user sees product names, prices, and in-stock status for the queried item and location
• user sees store names, addresses, and distances for location-based queries
• user sees cinema names, showtimes, and available seat counts for cinema requests
• user can copy a product ID, store code, or theater code from the results and use it in a follow-up query
• user reports "i found the item" or "which store should i go to" or "show me times for this movie"
• if no results, user says "try a different keyword" or "no stock in that area"
• if error, user sees the exact command that failed and the service health status, enabling