MCP use cases - rocketblue

Practical ways to automate AI visibility work with rocketblue MCP — digests, competitor monitoring, prompt hygiene, content pipelines, and more. Connect an assistant once, then ask it to run these workflows in natural language. Each use case lists the MCP tools involved so you can adapt the flow in Cursor, Claude Desktop, or any MCP client.

Weekly visibility digest

Goal: After each scheduled report, produce a short summary for Slack, email, or a Notion page. Ask the agent:

For brand BRAND_ID, get the latest run stats and top mentioned competitors.
Write a 5-bullet weekly digest: share of voice, biggest gains/losses vs prior run,
and one recommended action.

Tools: list_brand_runs → get_run_stats (current + previous run) → optional get_brand_last_results Goal: Flag when a competitor jumps in mention share on priority prompts or topics. Ask the agent:

Compare the last 4 runs for brand BRAND_ID. Highlight any competitor whose
topic-level mention share rose more than 5 points. List the prompts where they appear.

Tools: list_brand_runs → get_run_stats → get_run_results (only for flagged prompts)

Prompt portfolio hygiene

Goal: Keep the tracked prompt set aligned with real buyer questions and within quota. Ask the agent:

Check remaining prompt slots for brand BRAND_ID. Find prompts with zero brand
mentions across the latest run. Propose replacements that are neutral and
consideration-stage, then wait for my approval before deleting or creating.

Tools: get_brand_context → get_usage_guide → get_brand_last_results → delete_prompts / batch_create_topics_and_prompts

Citation gap and source map

Goal: See which domains LLMs cite for your category — and where you are missing. Ask the agent:

For the latest run of brand BRAND_ID, summarize cited domains from get_run_sources.
Call out domains that often cite competitors but not us, and suggest 3 outreach or
content angles.

Tools: list_brand_runs → get_run_sources → get_run_stats

AEO keyword mining from model search queries

Goal: Turn the queries models actually searched into a content and SEO backlog. Ask the agent:

Pull web search queries for brand BRAND_ID (ChatGPT and Perplexity only).
Cluster them into themes and output a prioritized brief list of article titles.

Tools: get_brand_web_search_queries (with llmNames) → optionally get_run_content_suggestions

Content pipeline from suggestions to drafts

Goal: Move from run suggestions into drafted articles your team can publish. Ask the agent:

List content suggestions from the latest run for brand BRAND_ID. Pick the top 3
by opportunity, then list any matching created-content drafts we already have.
If a draft exists, open it and summarize readiness to publish.

Tools: list_brand_runs → get_run_content_suggestions → list_brand_created_content → get_created_content

Publish-ready draft export

Goal: Hand CMS/docs a full markdown draft with schema hints. Ask the agent:

Fetch created content CONTENT_ID for brand BRAND_ID. Return the markdown body,
excerpt, suggested schema types, and research sources as a single publish package.

Tools: get_created_content

Perception score tracking

Goal: Monitor branded perception properties (trust, clarity, etc.) over time. Ask the agent:

Export perception prompt responses for brand BRAND_ID for the latest run and the
run before it. Table the score deltas and quote one representative LLM explanation
per property that dropped.

Tools: list_brand_runs → get_run_perception_prompts_responses (×2)

Agency multi-brand rollup

Goal: One assistant pass across every brand in the org for leadership reporting. Ask the agent:

List all brands. For each, pull brand context and latest run stats.
Produce a table: brand, plan slots remaining, last run date, main-brand mention rate,
top competitor. Flag brands with no run in 14+ days.

Tools: get_organization_context → list_brands → get_brand_context + list_brand_runs + get_run_stats per brand

Seed prompts from a research brief

Goal: Turn a GTM or category brief into a tracked topic/prompt set before the next report. Ask the agent:

Read this brief: [...]. Using get_usage_guide rules, propose topics and neutral
prompts for brand BRAND_ID. Check quota first. After I approve, create them.

Tools: get_usage_guide → get_brand_context → batch_create_topics_and_prompts

Tips for reliable automation

Always start with get_organization_context and get_brand_context so the agent knows plan limits and schedule.
Prefer get_run_stats for dashboards; use includeAnswer: true only when you need raw model text.
Writes land on the next scheduled report — digests should wait for a completed run, not fire immediately after prompt edits.
For bulk prompt cleanup, use delete_prompts once with all IDs.
Viewer/guest roles are read-only; write tools will fail for those accounts.

Connect an MCP client

Wire up Cursor or Claude Desktop.

Tools reference

Full tool catalog and REST mappings.

Best practices

Quotas, prompt style, and session flow.

REST API

Same data over HTTP for scripts and backends.

​Weekly visibility digest

​Competitor share-of-voice watch

​Prompt portfolio hygiene

​Citation gap and source map

​AEO keyword mining from model search queries

​Content pipeline from suggestions to drafts

​Publish-ready draft export

​Perception score tracking

​Agency multi-brand rollup

​Seed prompts from a research brief

​Tips for reliable automation

​Related