1
Sign in for the first time
Go to your tenant’s own address (for example
acme.app.us.dippr.ai) and sign in with the
password your deployment’s first administrator was given, or with Sign in with SSO if your
tenant connects an identity provider. Done looks like landing on the tenant’s home page rather
than a sign-in error. See Sign in to your tenant.2
Invite the first users
Go to Settings → Users, click Invite user, and send each teammate an invite link with
the role or roles they should join with. You need the Members · create privilege, which
account_admin holds. Done looks like each invitee opening their link, setting a password, and
signing in. See Invite a user.3
Assign roles
Every tenant starts with six built-in roles,
public through account_admin, each inheriting
the one below it. Review what each grants on Settings → Roles & permissions, and adjust a
member’s roles from the members list if the ones you invited them with aren’t quite right. Done
looks like every teammate holding a role that matches what they’ll actually do. See
Configure roles, row filters and column masks and
Roles and personas for which sections each persona needs.4
Decide where compute runs
Every tenant starts with a shared plane that Dippr operates, which is all most tenants ever
need. If your data has to stay inside your own cloud account, register your own plane instead
from Settings → Data planes, a
data_admin action. Done looks like the plane you intend to
use showing Connected. See Data planes overview and
Register or join a plane.5
Register the first integration and test it
Go to Data sources, open the Add integration tab, and click the card for the kind you
want (Postgres, ClickHouse, Snowflake, Google Sheets, DuckDB, or an Iceberg REST catalog); its
form opens already set to that kind. Fill in the connection details and click Connect: when
the source can be probed, Connect tests it first and won’t save if that test fails, unless you
explicitly check Register anyway, without a verified connection. Registering, editing and
testing need the
data_engineer role or higher.The Capabilities are set automatically from the source kind and shown read-only — you don’t
pick them. Every query-capable kind (Postgres, ClickHouse, Snowflake, Google Sheets, DuckDB, the
Iceberg REST catalog) is marked Queryable from engines by default, so it becomes a catalog on
your engines with no extra step. Done looks like the saved row on Connected; re-run the check
any time with Test connection on that row, which passes with the source’s own version string. See
Integrations overview and
Test connection and secrets.6
Create the first query engine and apply catalogs
Go to Settings → Query engines, click Create engine (a
data_admin action): name it,
pick a size, and register it on the same data plane as the source you connected — an engine
only serves sources on its own plane, so a mismatched plane yields an empty catalog set. Its first
catalog set — built from the queryable sources on that plane — applies
automatically. Later, whenever a source’s capability change
alters what an engine’s catalogs would be, the engine shows catalogs pending until you click
Apply, which restarts it. Done looks like the engine’s power state reading Ready with no
pending badge. See Create an engine and
Catalogs and Apply.7
Run the first query
Open Explore, open a new query tab, pick the engine and one of its catalogs (each queryable
source you connected on its plane), and press Run (
⌘↵). Any signed-in member can do this; there’s no separate permission to
unlock it. Done looks like rows coming back, filtered by whatever grants, row filters and column
masks apply to you. See SQL workbench.8
Turn on search indexing
Go to Settings → Search indexing and switch on indexing for each data plane you want
⌘K search to cover. Changing this needs permission to manage query engines; anyone else sees
the same page with the controls disabled. Done looks like the plane’s card showing indexing on
and, after the first crawl, a dip crawl status that reports a completed pass. See
Configure search indexing.9
Connect the first AI agent
From a terminal, install
dip and run dip login https://acme.app.us.dippr.ai. Signing in
registers your tenant’s MCP servers with the coding agents it finds on your machine as its last
step. Claude Code signs in with that same login, so while it is valid there is nothing more to approve
(if /mcp ever shows the server as needing authentication, run dip login again, then
start a new Claude Code conversation in the same folder and ask your question again). Done looks like claude mcp list showing the server as Connected, or a small SELECT 1 running successfully from the
agent. See Connect an AI agent.10
Optional: service users and API keys for automation
For a script, CI job, or an agent that shouldn’t run as a person, go to
Settings → Service users, create one, assign it roles the same way you would a member, and
mint an API key with only the capabilities it needs. You need the Members · create
privilege. Done looks like the key working from the caller it was minted for, with nothing
beyond the capabilities you ticked. See
Service users and API keys.
Related
Quickstart
The same path in one pass, screenshot by screenshot, for a single new member.
Roles and personas
Who on your team needs which of the sections above.