Get started

From nothing to a schema change applied in production: install tide, sign up for Atlantis Cloud, create an organisation, connect a repository, and apply your first entity.

1. Install tide

curl -fsSL https://releases.tryatlantis.dev/install.sh | sh

The script verifies the release checksum and installs into /usr/local/bin or ~/.local/bin, whichever is writable first. Confirm the install:

  ⊙
  tide  v0.4.0

2. Sign up and create an organisation

1. Sign up at Atlantis Cloud with your email, name, and a password, then open the verification link atlantis sends.
2. Set up two-factor authentication. Sign-in requires it, and your backup codes are shown exactly once — save them.
3. On the Organisations screen, create an organisation. The name is a DNS label: it starts with a letter or digit, then lowercase letters, digits, and hyphens, up to 63 characters. A few names — admin, api, atlantis, console, support, system, www — are reserved.
4. Wait for the organisation to report ready. Creating it provisions a dedicated database and server; the screen shows progress and links to the console when it is up.

3. Register a caller

A caller is a repository’s identity at atlantis (Manage members and callers). In your organisation’s console:

1. Open Callers and select Add caller.
2. Name the caller after the repository — api, for example — and enable mutate permission so it can plan and apply schema.

4. Connect the repository

In the repository root:

tide init --caller api
tide login

tide init writes tide.yaml, the repository’s only atlantis configuration file. tide login opens your browser: sign in, type the code the terminal shows, and approve the machine. The credential lands in ~/.atlantis/ and renews itself.

5. Declare an entity

Create schema.atl:

entity Note in app {
  id         bigint primary serial
  title      varchar(200) not null
  created_at timestamptz not null default now()
}

6. Plan and apply

tide plan
tide apply

tide apply runs the migration and records the new schema version. An additive change like this one applies unattended; a class that needs a decision waits in the console’s Approvals queue, and tide apply exits 2 to say the change is queued, not failed.

Verify

tide list

schema.atl

tide list prints every file the server holds for this caller — the one you just applied. The console’s Schema page shows the entity with its fields, and History shows the version your apply created.

Next steps

• Declare a custom query — SQL the typed surface doesn’t cover.
• Generate the typed client — set output_dir in tide.yaml and run tide generate.
• Add a new entity — the full modifier tour: references, soft delete, tenant partitioning.
• Set up CI — plan on pull requests, apply on merge, no repository secrets.