3. Going to production (later)#
Demo is the development environment — build and test everything there
first. When the store is ready for real data, ask the drop-core team for a
prod store-scoped access token (same store id), then add both lines to
.env:
DROP_ACCESS_TOKEN_PROD=… # the prod token
DROP_ALLOW_PROD=1 # explicit opt-in — without it prod stays unreachableBoth fences are required: a token without the opt-in (or the other way
round) leaves prod refused. The MCP server reads .env only at startup, so
restart your agent after editing it.
Keep DROP_ENV=demo — the workspace then still starts on Demo, and you
target prod per task: tell the agent to switch (useStore {env: "prod"},
and back when done) or aim a single call with the per-call env argument
every other tool accepts. Set DROP_ENV=prod only if prod should be the
workspace's day-to-day target.
The usual flow stays demo-first: build and verify a change on Demo, then
repeat it on prod on explicit instruction — the workspace's AGENTS.md
holds the agent to that. Prod's API base is https://core.inmake.dev (Demo:
https://demo-core.inmake.dev); a server of your own that reuses .env
(node --env-file=.env) picks its environment from the same file.