AI SEO Tracker MCP exposes seven tools. Names use snake_case with an aiseo_ prefix so they stay unique when other MCP servers are connected. The spec allows _ or -; most MCP tools use _.
Six tools are read-only.
aiseo_add_promptsis the only write. Tools do not send email, add competitors, or charge a card. Empty arrays mean the project has no data yet, not a failed connection.
Capabilities
- List projects the user can open, including
projectSlug. - Read scan state (
runningvsdone) without polling prompts. - Read citation sources (URLs and domains AI answers cite).
- Read tracked prompts with visibility status and per-model ranks.
- Read answer snippets for how a model describes a brand.
- Read competitor brands already tracked on the project.
- Add prompts to track, with the same plan limits as the website.
Tools
List accessible projects. Call this first for projectSlug.
Scan running vs done. Prefer this over polling prompts.
Citation URLs and domains. Optional domain filter.
Tracked prompts, status, ranks. Optional status.
How AI describes a brand. Optional brand.
Competitor brands for a project.
Add queries to track. Plan limits apply.
Call order
aiseo_list_projectsto getprojectSlug.aiseo_get_scanif a scan may still be running.donemeans no in-flight jobs, not complete coverage.aiseo_list_competitorsif you need rival names (do not invent them).aiseo_list_prompts,aiseo_list_sources, and/oraiseo_list_snippetswith that slug.aiseo_add_promptsonly for queries the user asked to track.
Do not guess slugs. If aiseo_list_projects is empty, the user needs a project at /new and a finished scan.
When to use which
| Tool | Use it when | Do not use it when |
|---|---|---|
aiseo_list_projects | You need a slug, or you are in a new chat | You already have a confirmed slug this session |
aiseo_get_scan | You need running vs done | You need ranks or citations |
aiseo_list_sources | You need sites AI already cites | You want ranks or quote-level copy |
aiseo_list_prompts | You need wins, gaps, or model ranks | You need the wording of an answer |
aiseo_list_snippets | You need how a model describes a brand | You only need a rank number |
aiseo_list_competitors | You need the tracked rival list | You are about to invent competitor names |
aiseo_add_prompts | The user asked to track specific queries | You are guessing prompts to “be helpful” |
Example session
A useful first prompt after MCP is connected:
Call
aiseo_list_projects. Use the first project. List missing prompts and the top citation domains. Do not invent ranks.
Typical tool sequence:
{}
aiseo_list_projects returns slugs. Then:
{
"projectSlug": "acme",
"status": "missing"
}
{
"projectSlug": "acme"
}
The agent should summarize from those payloads. Copy href fields for dashboard links. Do not invent paths. Setup encodes this order. Report writes the same brief as the Report page. Content coverage plans posts from missing prompts.
Errors that apply to every project tool
Wrong or inaccessible projectSlug:
Project "acme-typo" not found or you do not have access. Pick one of: acme, other-brand
No projects at all:
Project "acme" not found. You have no AI SEO Tracker projects yet - create one at /new.
Empty results include dataStatus: ok, no_data (nothing in the project yet), or empty_filter (filter matched nothing). That is not a 401. See Connect MCP.
Auth failures return HTTP 401 with error: "agent_auth_required". See Auth and Troubleshooting.