generated: '2026-08-05'
method: searched
source:
  - https://dev.tapcart.com/docs/mock-data-vs-hooks
  - https://dev.tapcart.com/docs/blocks
  - https://dev.tapcart.com/docs/mcp
description: >-
  Tapcart publishes NO hosted sandbox, no test-mode key prefix, and no
  test-vs-live account separation. What it does publish is a local development
  surface, and this artifact records that honestly rather than inventing a
  sandbox that does not exist. Three things stand in for one: a local hot-reload
  dev server, a developer-controlled mockData.json fixture file, and a
  plan/apply gate on the MCP write tools. Crucially, the docs state the limit
  explicitly — mockData.json does not mock network-backed hooks, so any block
  using useCollection() hits the real Tapcart APIs with real app context even
  in local development.
test_live_separation:
  supported: false
  key_prefixes: null
  test_mode: null
  note: >-
    There is one set of credentials. The CLI's Auth0 login and the app's API
    key are production credentials; there is no test key, test app, or sandbox
    tenant documented.
local_development:
  dev_server:
    command: tapcart dev
    modes: [block, component, layout]
    flags: [target, port, verbose]
    description: >-
      Hot-reload local server for previewing a block, component or layout. Run
      with no target to pick one from a browser picker.
    mcp_note: >-
      The MCP server does not start the dev server; tapcart_dev_instructions
      returns the command for a human or agent to run.
  fixtures:
    file: mockData.json
    location: project root (exactly one per project)
    created_by: "tapcart project create / tapcart block create (never overwrites an existing file)"
    read_by: the local dev server, surfaced through useVariables()
    overrides: [cart, customer, product, collection, device, and other app variables]
    published_example:
      note: Transcribed verbatim from the docs; identifiers are placeholders in the source.
      value: |
        {
          "customer": { "id": "<UUID>", "firstName": "Taylor" },
          "device": { "locale": "en-US" },
          "cart": { "items": [] }
        }
    limitation: >-
      "mockData.json does not mock or override network-backed hook results." A
      block that uses useCollection() or any other network-backed hook still
      fetches real data from Tapcart APIs, and local preview may require a
      valid project/app context.
write_safety:
  draft_then_live:
    description: >-
      Pushing a block or component stores it as a new version on the server as
      a DRAFT; it does not go live until promoted. `--live`/`-l` on push, or a
      versions_set call, promotes it. This draft stage is the closest thing to
      a staging environment Tapcart offers.
  mcp_mode_gate:
    description: >-
      Every remote-write MCP tool defaults to "plan" (no remote write) and
      requires "mode": "apply" to actually write. The published agent skills
      describe the same flow as plan then confirm then apply.
    tools: [tapcart_blocks_push, tapcart_block_versions_set, tapcart_components_push, tapcart_component_versions_set, tapcart_dependencies_pushRemote, tapcart_lint]
  no_local_tracking_warning: >-
    Documented hazard: "The Tapcart CLI does not track any local changes.
    Pulling a component will overwrite any changes with the currently live
    version."
webhook_testing:
  documented: false
  note: >-
    No webhook replay, event trigger, test-event, or local-forwarding tool is
    published for the clickstream webhook. A merchant must point the dashboard
    at a reachable HTTPS endpoint and generate real app activity.
gaps:
  - No hosted sandbox or test tenant.
  - No test-mode API keys or key prefixes.
  - No webhook event simulator or replay.
  - Network-backed hooks cannot be mocked locally.