# How SERPclimber works

SERPclimber is an SEO workspace organized around **projects**. Each project is one website: its goal, business context, connected data, allowed scope, and autonomy policy.

Free projects can collect connected first-party data. Managed projects also run the engine's continuous loop:

1. **Observe** — crawl the site, check how AI engines answer questions in your space, and add Search Console or GA4 measurements when connected.
2. **Audit** — turn those observations into evidence-backed issues and opportunities, grouped by department.
3. **Prioritize** — re-rank the backlog against the project goal every hour and promote only the most useful findings into Tasks.
4. **Work** — execute approved tasks through the engine, you, or an external agent connected over MCP.
5. **Verify** — check the live result and measure the outcome when a measurement window is useful.
6. **Report** — write weekly or monthly digests of what changed, what improved, and what comes next.

## The workspace

Projects added from Search Console open on **Performance → SEO**; manually added projects open on **Tasks**. Each has five sections: **Performance** (with **SEO**, **GA4**, **AI answers**, **Crawlers**, and **Authority** tabs), **Tasks**, **Audit**, **Reports**, and **Settings**. Every task has a project-wide number such as `#42` and its own thread, where evidence, plans, approvals, receipts, and your comments live together.

The assistant entry point appears throughout the workspace: on its own page, in a pane pinned beside any page, and behind **Explain with AI** on Performance cards. Replies require a paid plan, and project-specific analysis requires a managed site; free projects show an upgrade path. Once managed, ask it what needs you, why a task exists, or what a chart means. See [Chat](/documentation/chat/overview).

## How much the engine owns

- **Audit only / MCP** — the engine observes, files findings, and plans. You or your external agent perform every change.
- **Copilot** — the engine performs low-risk, reversible work and asks before risky, external, or paid actions. New projects start here.
- **Full** — the engine performs everything its policy allows, including paid evidence within budget.
- **Custom** — set a default rule and override it per department or action class.

Start by [connecting your data](/documentation/getting-started/connect-your-data) and adding a free site from your Search Console catalog, or add a site manually. Make a site managed when you want automated audits and paid research.