MCP Connector | AI Visibility Portal Help

Q: Can the assistant publish a result or start a scan?

No. The connector has no tool that starts scans, publishes releases, manages access keys, or starts paid reruns.

Connect your assistant

In your assistant, open Settings, then Connectors, and choose the option to add a custom connector.
Paste this connector address and confirm:

https://mcp.visibility.creative-data-engineers.com/mcp

Sign in with your existing portal email and password.
If you belong to more than one account, select the account you want to connect.
Review the connection and approve it.

The connection is tied to your current account membership. If that membership becomes inactive or the connection is revoked, access stops immediately.

Settings drawer open to Claude and MCP connector setup — The MCP settings provide the connector address and setup steps for a supported assistant.

What the assistant can do

The connector exposes 15 tools. Eight work with published results and reversible recommendation marks. Seven work with actor-owned onboarding state and transactions. Onboarding writes use current versions, an idempotency_key, and the connected member's identity. Successful writes return stored evidence such as a digest, version, actor display snapshot, and safe job state.

The boundary is fixed.The connector cannot approve your questions, start a scan, publish a release, review a service tier, manage access keys, or start a paid rerun. Question approval remains a portal action.

Resources and guided prompts

Protocol version 2025-03-26 provides standard resources/list and resources/read methods. They expose onboarding status, current brand authority, the current question proposal, and onboarding methodology. Onboarding status and current brand authority can also return the exact current preliminary preview. Current resources never include credentials, raw GSC errors, internal job payloads, or another member's notification preferences.

Use prompts/list and prompts/get to discover three guided workflows:

review_brand_facts checks the current facts before saving or confirming them.
prepare_measurement_setup prepares a complete setup and explains its acknowledgement.
review_question_set reviews the current proposal and prepares feedback without approving it.

Preliminary preview

The label Preliminary: one of three scheduled samples identifies the verified Monday sample. After Wednesday passes verification, the label changes to Preliminary: two of three scheduled samples. This second preview contains Monday and Wednesday. Friday is still outstanding.

Both preview versions are available through get_onboarding_status and the matching status and authority resources. The MCP payload matches the portal payload.

The three-sample baseline becomes available after Friday passes verification. Until then, deltas, trends, recommendations, and any measured or live status remain unavailable.

Fallback evidence gives the preliminary preview reduced confidence. The returned confidence limitation names the hypothesis evidence so an assistant can explain the boundary.

Published-result tools

Published-data tools read only the current published release for a brand in the connected account.

`list_brands`

Inputs: None.
Returns: Published brands with slug, display name, domain, data state, current release version, and publication date.

`get_brand_overview`

Inputs: brand_slug: the short identifier returned by list_brands.
Returns: The brand overview from the current published release.

`get_measured`

Inputs: brand_slug and section: search, ai, fanout, or readiness.
Returns: The requested measured section.

`list_documents`

Inputs: brand_slug.
Returns: Current-release document metadata.

`get_document`

Inputs: brand_slug and the exact file_name.
Returns: The document text and a truncation indicator for very large files.

`get_schema_drafts`

Inputs: brand_slug.
Returns: Current-release JSON-LD drafts.

`get_recommendations`

Inputs: brand_slug and optional status.
Returns: Current published recommendations with reversible client marks applied.

`update_recommendation_status`

Inputs: brand_slug, item identifier, action, and optional reason.
Returns: The effective recommendation status.

Permission: Marks an item done or dismissed, or restores it. This is not an approval action.

Actor-owned onboarding tools

These tools read or update current onboarding authority. Inputs go directly into actor-attributed, versioned transactions. They do not enter a CDE approval queue.

`get_onboarding_status`

Inputs: brand_slug.
Returns: Authority versions and digests, your notification preferences, current proposal state, and safe onboarding job states.

`save_brand_facts`

Inputs: brand_slug, a multi-field patch, expected_profile_version, and idempotency_key.
Returns: Stored status, changed fields, profile version, request digest, actor evidence, and safe job state.

`confirm_brand_facts`

Inputs: brand_slug, expected_profile_version, and idempotency_key.
Returns: The confirmed version, stored request digest, actor evidence, and safe job state.

`apply_measurement_setup`

Inputs: brand_slug, complete setup, expected_measurement_version, acknowledged set to true, and idempotency_key.
Returns: The new version, stored profile and acknowledgement digests, actor evidence, and safe job state.

Acknowledgement: The change creates a trend boundary, requires question-bank regeneration, joins the next included monthly run, and does not start an immediate paid run.

`set_notification_preferences`

Inputs: Purpose, enabled or disabled value, expected_preference_version, confirmed set to true, and idempotency_key.
Returns: Your stored preference version, request digest, and actor evidence.

`set_gsc_unavailable`

Inputs: brand_slug, expected_gsc_version, acknowledged set to true, and idempotency_key.
Returns: The fallback state version, state and acknowledgement digests, actor evidence, and safe job state.

Acknowledgement: The fallback uses no observed GSC demand, has reduced confidence, and later GSC evidence invalidates an unconsumed proposal.

`submit_question_feedback`

Inputs: brand_slug, feedback_kind, question_ref, proposed_question, rationale, and idempotency_key.
Returns: The feedback identifier, stored request digest, actor evidence, and safe job state.

Boundary: Feedback can regenerate a proposal. It cannot approve or directly replace the question bank.

Ask in plain language

You do not need to remember the method names. Ask the assistant to review your current brand facts, prepare measurement setup, explain onboarding status, review the proposed questions, summarize readiness, or download the latest AI brief. The assistant should ask for explicit confirmation before a fact confirmation or required acknowledgement.

Frequently asked questions

Which account can the assistant reach?

The connection is limited to the portal account approved during sign-in.

Can the assistant publish a result or start a scan?

No. It cannot start scans, publish releases, manage access keys, or start paid reruns.

Can the assistant approve your questions?

No. Question approval remains a portal action. The assistant can review the proposal and submit feedback.

Can the assistant update onboarding inputs?

An actor-owned connection can save and confirm brand facts, apply measurement setup, set notification preferences, acknowledge the GSC fallback, and submit question feedback.

← Back to all help guides