Is Cursor Pro still a plan, and what security work fits it?

As of our fetch on 18 October 2026, Cursor's pricing page no longer lists a plan called Pro. It shows Free, Plus, Premium, Super and Ultra, plus Teams and Enterprise, and the docs call Pro, Pro+ and Ultra legacy plans. The practical rule: put cheap, repeatable checks outside the metered agent and save agent usage for work that needs judgment.

This post uses only published prices and limits from Cursor's own pages on that date. Prices are subject to change, as the page says, so check before you decide. We make no claims about what any plan costs Cursor to run.

What do the plans look like today?

From the pricing page and the plan docs, fetched 18 October 2026:

• Free: $0, with limited agent and chat requests.
• Plus: $8 per month, described as an introductory plan.
• Premium: $30 per month, with 4x the usage of Plus.
• Super: $100 per month, 15x Plus.
• Ultra: $200 per month, 50x Plus.
• Teams: from $30 per user per month. Enterprise: custom pricing.

The page says the allowance refreshes weekly. It also says that on Pro, Pro+, Ultra, Start or a Teams Standard or Premium seat, you are on legacy plans that keep working for existing customers, with two monthly usage pools and on-demand usage. The docs we read give no price or usage figure for legacy Pro, so we will not state one.

The pricing page also described some plan features as tied to particular model providers, and the Plus plan as limited to a specific provider's models. Read the live page before you assume which models your plan covers.

Which security work uses agent usage?

Anything where you ask an AI agent to read code and reason: reviewing a pull request for authorization flaws, explaining a finding, writing a fix across several files. These calls are the metered part, and a long agent session on a large repository can use a lot.

Which security work does not need the agent?

Work that is lookup or rule based should not burn your allowance:

• Dependency scanning. Matching package versions to advisories is a deterministic job. Safeguard's IDE extensions work in Cursor through Open VSX, and the CLI scans from the terminal, neither of which depends on your Cursor plan.
• Secret detection. Pattern based scanning in a pre-commit hook or CI.
• Policy gates. Blocking a merge on a rule, in CI.
• MCP queries. Safeguard's MCP server offers unlimited tool calls on every plan. Your Cursor plan still governs the model usage around those calls, so a long conversation with many lookups still costs agent usage on the Cursor side.

How do you split the work?

A workable split for a small team:

1. In the editor, always on: a dependency and secrets scan on save or commit. No agent needed.
2. In CI, every pull request: the full scan with a policy gate.
3. In the agent, on demand: triage of the five findings the scan could not rank, and fixes that span files.
4. Weekly: a human reads the agent's proposed diffs from the week and samples them for errors.

What should you watch out for?

Agent generated code needs the same scanning as any code. Models can suggest packages that do not exist, which an attacker could register, so verify any new dependency an agent adds against your inventory and scanner before merging. Do not give an agent credentials broader than the task needs, and keep secrets out of prompts.

What to do

1. Open the live pricing page and note your plan, the date and the usage terms.
2. Check whether your seat is a legacy plan, and read what that means for usage pools.
3. List the security checks you run and mark each as lookup or reasoning.
4. Move lookup checks to the editor extension, pre-commit and CI.
5. Reserve agent sessions for reasoning tasks with a clear scope.
6. Review your usage weekly for the first month and adjust.

Quick check

Sources

• Cursor pricing page, fetched 18 October 2026
• Cursor plan documentation, fetched 18 October 2026

Last reviewed: 18 October 2026.