FIELD_MANUAL / LIVEREVISIONv1.3 · 2026.10.04Updated 2026-10-04CHAPTERS12 FILEDCH_05INTEGRATION

[ CH_05.03 / FIELD_MANUAL ]

MCP server

[ CHAPTERS ]

[ CH_05.03 / MCP_SERVER ]

Drive VectraSEO from an AI assistant

VectraSEO ships a Model Context Protocol server, so an assistant can create, onboard and manage projects directly. Same account key, same permissions as the Provisioning API.

> endpoint

https://api.vectraseo.com/mcp

Connect

claude-code.sh

claude mcp add --transport http vectraseo \
  https://api.vectraseo.com/mcp \
  --header "x-api-key: vsak_..."

Any client that speaks Streamable HTTP can point at that URL with an x-api-key (or Authorization: Bearer) header. For clients that can only launch a stdio subprocess, run npx -y @vectraseo/mcp with VECTRASEO_API_KEY set — a zero-dependency bridge to the same endpoint.

Tools

TOOL

DOES

whoami

Identify the key: account, plan, project count. Call this first.

list_projects

Every project on the account.

get_project

One project's full configuration.

onboard_project

Create AND fully configure a project in one call.

get_onboarding_status

Poll an onboarding run, step by step.

update_project

Edit any project — settings, brief, generation policy.

list_competitors

Competitors tracked for a project.

add_competitor

Attach a competitor URL.

delete_competitor

Stop tracking a competitor.

list_monitors

A project's site-health monitors.

create_monitor

Create a monitor over a sitemap.

update_monitor

Edit a monitor — most usefully its alert_emails.

run_site_scan

Scan now instead of waiting for the schedule.

get_recommendations

Every current SEO/AEO/Search-Console fix, with the concrete change.

get_recommendation_detail

Open one recommendation: candidate titles, metas and links. Paid plans.

apply_recommendation

Apply a chosen fix to a post. Republishes to the live site.

apply_all_recommendations

Fix everything applicable in one job. Call with dry_run first.

dismiss_recommendation

Hide one recommendation until its Search Console numbers change.

list_gaps

Suggested topics, paginated. Where a gap_id comes from.

reject_gap

Mark a topic off-target. Kept as a negative example.

delete_gap

Delete a suggested topic outright. Prefer reject_gap.

list_visibility_monitors

AI-visibility monitors, each with its prompt list.

get_visibility_monitor

One visibility monitor: every prompt, text and intent.

update_visibility_monitor

Edit a visibility monitor — most usefully its prompts.

get_indexing_status

Google Indexing API state, last submission and last error.

set_schedule

Recurring content schedule.

trigger_content_run

Run the full content pipeline. Spends budget.

get_job

Poll any asynchronous job.

list_posts

A project's blog posts, paginated.

publish_post

Publish or republish. Writes to the live site.

Safety

Tool calls run through the same REST API with the same credential, so an assistant can never do anything the API would refuse — the scope check, plan gate and ownership check all run in their usual place. Tools that spend budget or write to your live site are flagged destructiveHint so your client can prompt first, and the server tells the assistant to confirm with you before calling them. No tool can delete a project, and no tool can manage API keys — including its own, so a leaked key cannot perpetuate itself.

[ END_OF_MANUAL ]

[ NEED_HELP? ]

Can't find what
you're looking for?

Reach out to the team. We answer every email and we read every bug report.

contact_operator → or email support@vectraseo.com