# CLI overview (/docs/cli/overview)



The Pollaris CLI lets terminal-capable AI agents create forms without asking you to navigate the Pollaris dashboard.

Ask your agent:

```text
Create a library feedback form via Pollaris.
```

The agent runs Pollaris, which returns a secure authorization link. Open the link, sign in or create an account, choose a workspace when needed, and approve access. The agent runs the returned continuation command and returns a published URL such as:

```text
https://forms.pollaris.cc/f/example
```

For a new account, Pollaris can create a personal workspace during authorization. Email verification returns you to the same authorization request. The normal flow does not require an API key, dashboard, editor, or Publish button.

## What happens after approval [#what-happens-after-approval]

Pollaris grants the requesting CLI profile access to one workspace. The backend generates the form, validates every question and branch, saves it, publishes an immutable version, and gives the agent its public URL.

The grant is limited to the permissions shown on the authorization page. It can be revoked with `pollaris auth logout` or through the connection command described in [Agent connections](agent-connections.md).

Your agent host must have the Pollaris CLI installed or a Pollaris MCP connection configured. Mentioning Pollaris in a prompt cannot install software when the host does not permit terminal commands or integrations.

## Drafts and published forms [#drafts-and-published-forms]

The high-level create command publishes by default because its goal is to return a usable link. Ask for a draft or use `--draft-only` when you want to review the form first:

```bash
pollaris forms create \
  --prompt "Create a library feedback form" \
  --draft-only \
  --json
```

Continue with [Installation](install.md), then [Create a form](create-forms.md).