whats_next

Cost: Free (0 credits).

whats_next is the guide for non-experts. It looks at where a project stands — whether it has been crawled, whether Google Search Console is connected, whether you have pulled performance data — and tells you the single best next step, a short reason, and the two or three steps that come after.

What it does

Pass a project_id to route that project; omit it and it routes your only project, or lists them and asks which one if you track several. It reads the project's state through the same tenant-scoped data the tools use — your latest crawl_site and pull_gsc_data runs, whether Search Console is connected, whether that connection is still alive, and whether the domain resolves — then walks a ladder, taking the first rung that applies:

• No project yet → run setup_project.
• The domain does not resolve → nothing to measure yet. This rung names no paid tool at all, because every recommendation below it would be work against a host that is not there.
• No crawl yet → run crawl_site (works without Search Console).
• Crawl ready, Search Console never used → run the audits: audit_onpage, audit_tech, audit_schema. Connecting Search Console with connect_gsc is optional and never a barrier.
• Old Search Console data but no live connection → connect_gsc, which is free. Frozen rows cannot be refreshed, so nothing that reads them is offered.
• Connected, but the credential is dead → connect_gsc again, and the discovery tools are deliberately withheld: they read a pull this project cannot take.
• Connected, but no property is mapped → list_gsc_properties, then track_gsc_property. Both are free. The Google account works, so this rung does not send you round another OAuth loop — what is missing is the mapping, and until it exists a Search Console pull cannot run.
• Connected, nothing pulled → run pull_gsc_data.
• A pull or a crawl has gone stale → refresh that one first, so the numbers describe the site as it is now.
• Everything fresh, but nothing analysed yet → find_quick_wins first: it reads the Search Console data you already pulled and names the pages closest to moving up. A report is worth generating once there are findings to put in it.
• Everything fresh → you're all set: generate_report for a shareable summary, and the monthly-routine prompt to keep it up to date.

"Rows were pulled once" is not "the link is live"

The two used to be the same signal, and the difference is what the middle of that ladder is for. A succeeded pull is a fact about the past: it outlives a disconnect, an unmapped property and a deleted Google account. A project holding one such pull and no connection was told it was all set and pointed at a paid report — while the free reconnection that was the real next step went unmentioned.

So liveness is now read separately, and a project that cannot refresh its Search Console data is never called all set, however fresh the frozen rows look.

Every step says what it costs

The primary step and each line of what follows carry their own price — free, or the credits that tool charges. This tool is for the reader who does not already know the price list, and a list that named a free step and a paid one in the same breath left them nothing to choose with.

The figures are read from the same signed cost table the tools charge from, never restated here, and a tool priced per unit shows the range a call can really cost rather than a unit price no call ever pays. A step that is not a priced tool — a prompt, or a note about coming back later — shows no price rather than a guessed one.

Example

Ask your MCP client in plain language:

What should I do next with my example.com project?

Input

Returns

One clear next step for the project, a short reason, and the next two or three steps — all in plain language, naming the exact tools to run and what each one costs. An archived project is refused rather than routed, and a project_id that is not yours is reported exactly like an id that does not exist.