Skip to content

Sign in to Cursor and choose a plan

Cursor is installed and budgetcli is open in it, but on a fresh machine the editor doesn’t yet know who you are — the Agent sidebar will prompt you to sign in before it does anything useful. This is the shortest consequential lesson in the chapter: the click itself is trivial, but it’s also the moment Cursor switches on a privacy setting that decides what leaves your machine, and that’s worth thirty seconds of attention before the agent starts reading code you didn’t write.

You don’t manage a password inside Cursor. On first run the app — and the Agent sidebar specifically — surfaces a Sign in prompt; clicking it opens your browser, you authorize the session against your Cursor account, and you land back in the editor logged in. If you don’t already have an account you create one in that same browser step. There’s nothing to copy-paste and no key to manage for interactive work at your own desk — the browser round-trip is the whole flow.

Your account isn’t only a login. It’s the thing that carries your plan and its usage allowance, so the same sign-in that unlocks the editor is also what meters the agent. The previous lesson laid out the tiers — a free Hobby tier up through Pro and Ultra, with Teams and Enterprise layering on admin controls — and the free Hobby tier is enough to follow this chapter. The point worth holding onto: the account is the unit everything attaches to — your plan, your usage pool, your settings, and the privacy posture below all hang off this one sign-in.

You’ll meet Cursor’s terminal agent properly in a much later chapter, but it’s worth knowing now that it shares this account — you authenticate it once, the same way:

agent login

That opens the same browser flow and stores the credential locally; agent status confirms who you’re signed in as, and agent logout clears it. The binary is agent today; cursor-agent is a legacy alias for the same tool, so older docs and scripts that say cursor-agent still work — the CLI chapter writes it that way on purpose to keep it unambiguous, but agent is the current invocation.

For unattended use — CI, a scheduled job, anything with no human to click a browser prompt — the browser path is exactly wrong, the same way it is for any agent. There the CLI authenticates with a user API key instead, generated from the Cursor dashboard and passed as CURSOR_API_KEY (or --api-key). You don’t need it for anything this hour — just file away that the interactive door (browser) and the automation door (API key) are different doors, and the headless chapter walks through the second one.

Here’s the part that earns this lesson its thirty seconds. Tied to your account is Privacy Mode, a setting that governs what Cursor is allowed to retain when your code passes through its servers. It’s a real toggle — settable per account, and enforceable across a whole team or organization — and the trade-off is concrete:

  • With Privacy Mode on, Cursor stores no plaintext of your code on its servers and never uses it as training data; any file contents that transit a request are transient.
  • Codebase indexing is the feature most affected by that toggle. To answer questions about a project as a whole, Cursor can build a semantic index: it uploads your code in small chunks to compute embeddings, discards the plaintext after the request, and keeps the embeddings plus metadata (hashes, obfuscated file paths) so the agent can later search the codebase by meaning rather than by you naming files.

For budgetcli — a personal finance API with your own money in it — that distinction is not academic, which is exactly why it’s surfaced this early. You don’t have to decide the indexing question right now; the date-bug fix this chapter ends on doesn’t require a whole-repo index. The mechanics of how Cursor pulls the right context into a turn — indexing, what gets uploaded, and how to tune or disable it — get their own treatment in the context chapter later in the course. For this hour, just know the toggle exists, know which way it leans for code you care about, and leave it at its default until that chapter gives you the full picture.

One quick check before you move on. Open the Agent sidebar, type a harmless question about the repo — “what does budgetcli do?” — and send it. A coherent answer that references actual files means three things at once: you’re signed in, your plan’s agent access is live, and Cursor can read your workspace.

You’re signed in, the agent answers, and you know what your account switched on. Next, the part that trips people up: the two AI surfaces you just met — the Agent sidebar and Tab — aren’t interchangeable, and reaching for the wrong one is the most common early mistake. Next: Agent sidebar vs Tab.