Skip to main content
This quick start gives you an overview-level path from zero to your first useful conversation with Conductor’s MCP. It is intentionally LLM-agnostic—the same five steps apply whichever MCP-capable AI assistant you connect through. If any term is unfamiliar, skim Introduction to Conductor’s AI Connectors and What is an MCP tool? before you start.
Want a more in-depth introduction to Conductor’s MCP and all you can do with Conductor data in your AI tools? Conductor customers can take our Community tutorial: Power Your AI Tools with Conductor Data

​
Getting Started with Conductor’s MCP

1

Configure an MCP connection with your AI tool of choice

Before you can ask anything, your AI assistant needs a connection to Conductor’s MCP server. Each LLM does this a little differently—the underlying protocol is the same, but the UI and plan requirements vary.Pick the assistant you use and follow its setup article:At a high level, each guide walks you through:
  1. Verifying your plan and permissions
  2. Adding a connector pointed at Conductor’s MCP URL
  3. Authenticating with your Conductor account.
Once you’ve enabled the connector, your assistant can call Conductor’s MCP tools on your behalf.
Need approval from your IT team?
2

Go shopping for skills

You don’t have to invent every prompt from scratch. Conductor maintains a library of skills—pre-built, tested prompt templates that take advantage of MCP tools to deliver a specific outcome (an audit, a competitive comparison, a board-ready summary, and more).Browse the Template Gallery to find skills that match the work you want to do. Useful starting points include:
  • Introduction to using skills covers how skills work and how to run one with your assistant.
  • Teach your LLM Conductor gives your assistant the vocabulary and context it needs to use Conductor data well.
  • The gallery itself, filtered by persona or use case to narrow down.
Treat skills as a launchpad. Run one to see the shape of the answer, then adapt your prompt to investigate different topics, competitors, and time ranges.
3

Identify the question you want to ask

The MCP is most useful when you bring a real question. A vague “how are we doing?” produces vague answers; a specific question routes to the right tool and returns something you can act on.Start from the data that’s available and pick a slice that matters this week. Common starting points include:
  • AI search — brand mentions: Which topics or queries drive the most brand mentions for you? Where do competitors out-mention you? How is share of voice trending?
  • AI search — citations: Which of your pages and domains earn the most citations? Which topics mention you but don’t cite you as a source?
  • AI search — sentiment: How do AI responses talk about your brand—positive, neutral, negative—and which topics carry the most risk?
  • Traditional search — rankings and visibility: Which tracked keywords moved most over the last quarter? Which keyword groups or locations are driving the change?
  • Traditional search — seasonality, SERP features, and competitive rankings: When does demand peak for your priority keywords? Which SERP features dominate the page? Where are specific competitors pulling ahead?
  • Account configuration: What topics, prompts, brands, competitors, personas, intents, and locales are you tracking today—and what’s missing?
See What can I do with Conductor’s MCP? for more robust examples of each.
Use vocabulary Conductor already uses—brand mentions, citations, topics, queries, keywords, keyword groups, rankings, SERP features—instead of vague words like “visibility.” Specific language routes to the right tool and produces clearer answers.
4

Have your conversation about your data

Once the connector is live and you have a question in hand, treat the assistant like a research partner—not a search box.A productive flow usually looks like this:
  1. Orient the assistant. Ask it to confirm which Conductor account it’s connected to and what tools and time ranges are available. A quick prompt like “Show me which Conductor accounts I have access to and the tools you can call on my behalf” prevents wasted turns later.
  2. Ask your starting question. Be specific about the scope of your research—topic, competitor, keyword group, time range, or persona. If you are using a skill, invoke it with a slash command.
  3. Read the answer critically. Check that the assistant used the right tool and the right data (the account, time range, and entities you expected). If it summarized without calling a tool, ask it to query Conductor explicitly.
  4. Dig deeper. Follow up with “why” and “what next” questions. For example: “Which pages drive that citation share?” → “Which of those pages have lost citations in the last 8 weeks?” → “What else can you tell me about those pages’ performance?”
  5. Turn insight into action. Ask the assistant to summarize the takeaways, draft a recommendation, or produce a board-ready narrative. Pair this with skills like Build a board-ready AI search deck or Build a data-backed content strategy.
If the assistant gets stuck, narrow the question—shrink the time range, name a specific competitor, or pick one topic—and try again. The MCP rewards precision.
5

Turn what you found into content

The first four steps only read your data—they tell you what’s happening. The Content tools are where you act on it. They run the Writing Assistant workflow in the same conversation, so the page you just diagnosed becomes the page you score, brief, and create or update without switching tools.What you found in the previous step might point to one of these follow-ups:
  • A page that gets mentioned but rarely cited. Create a brief for that URL, score its content, revise the content, and then score it again to confirm the revision helped.
  • A topic where competitors out-mention you and you have no page. Create a brief for new content about that topic and generate a first draft.
  • Keywords or citations that slipped last quarter. Create a brief (or open an existing draft) for the live page that’s losing ground, then work through the score’s prioritized improvements in order.
  • Research scattered across transcripts, notes, and reports. Combine it into a Knowledge Source so future content is grounded in your own material rather than the model’s general knowledge.
A typical pass looks like this:
  1. Start from the page or topic you just investigated. Ask for a content brief, pointing at the live URL you want to improve or naming the topic you want to cover. The brief describes what a well-optimized version looks like, and a brief built from an existing URL arrives with that page’s score already attached.
  2. Review the brief before anything is written. Read it back and tailor it to your goals before generating any changes.
  3. Draft or revise against it. Attach a Content Profile so the output carries your brand voice and audience instead of a generic one.
  4. Finish the piece. Ask for title and meta description options, then for internal link suggestions. Both give you candidates to choose from; nothing is applied until you pick.
  5. Score it again and open it in Conductor. Compare against the baseline from step 1, then follow the link to the draft to keep working in the platform.
Content tools are in beta and are not currently included in the native connectors. Set up a custom connection to use them.Creating a brief uses a draft from your organization’s allotment. However, using and revising that same draft carries through every later step at no extra cost.
Pair this step with a skill rather than starting cold. Build a data-backed content strategy takes a topic and returns the tracked prompts behind it, which competitors are winning citations, and the top-cited competitor URLs with why they’re winning.Hand it a list of the URLs you have already published on that topic—a Conductor Pages export, a Conductor Monitoring crawl, or a sitemap—and it marks each gap Create or Optimize: topics you genuinely lack, versus pages you already have that simply aren’t being cited. That tells you which kind of brief to open this step with.

​
What to keep in mind

  • Validate important decisions—the assistant summarizes and explores, it does not replace your judgment.
  • Know which tools write. The analysis tools only read your data. The content creation tools create and change content in your account and can consume allotted drafts.
  • Iterate. Your first prompt is rarely your best. Refine it based on what the tools return.
  • Ask for help. Your Conductor team can share updated prompt examples and recommend skills that match your program.

​
Next steps