Skip to main content
Here are frequently asked questions about Conductor’s MCP server and its various connectors.

​
About Conductor’s MCP Server

Don’t worry! While it’s easy to get caught up in the terminology, this feature is super straightforward to set up and use. You’ll just find the connector in your AI account, enter your Conductor credentials, and you’re ready to go.An MCP (model context protocol) server is just a technology that lets existing AI models (like ChatGPT or Claude) connect to data from a third-party. Conductor’s MCP server connects AI models to your Conductor data, so you can query the AI model about your data.
Standard LLMs generate answers based on training data, which is often outdated or prone to “hallucinations” (guessing). Conductor’s connector uses a “split reasoning” architecture: LLMs understand the user’s intent and handle the conversation, but Conductor’s API provides the data and facts. This ensures you get verified, accurate intelligence rather than a best guess.
“Local MCP” means the connection runs directly on your personal computer, whereas “remote MCP” runs in the cloud as a direct server-to-server connection. Conductor has architected our platform to support both. While we are launching our Remote MCP first to power this integration with ChatGPT, this infrastructure is universal—allowing us to easily expand our verified connector to other LLMs and interfaces (e.g., Copilot, Gemini, etc.) in the future.
Conductor’s API is the core infrastructure that allows you to request data programmatically for any use case—including BI dashboards, custom applications, and automated reporting. The MCP Server is a specialized connector built on top of the API that formats this data specifically for AI agents like ChatGPT and Claude. You get access to the power of both.
Both pull from the same Conductor data—the difference is what you get back and what you plan to do with it. Use the MCP when you want an answer: you’re asking a question in plain language and want a clear, insightful response, ideal for exploring, spot-checking, and “just tell me what’s going on.” Use the API when you want the raw data to build something, like piping large exports into a BI dashboard (Tableau, Power BI, Looker), a data warehouse, or a report that refreshes automatically. That’s a plumbing job and generally needs a developer or data team to set up and maintain.If you’re not sure which to use, ask yourself: are you trying to understand something (MCP), or feed another tool (API)?
Not with Conductor’s MCP. However, you can with Conductor’s API. Conductor doesn’t offer a native connector for BigQuery, Snowflake, Databricks, or other data warehouses. Instead, your data or engineering team uses the Data API to request Conductor data and load it into your warehouse on a schedule. Data API requests are asynchronous: you submit a query, check back until the results are ready, and then page through them. Once the data is in your warehouse, any BI tool connected to it—like Looker, Tableau, or Power BI—can report on it.For Conductor Monitoring data, use the Conductor Monitoring Reporting API or the Looker Studio connector.

​
Data & Capabilities

Check out our article about the available tools you can call with your LLM.
Most of it can’t. The AI search, traditional search, and account configuration tools are read-only—they query and analyze your Conductor data without changing it.The exception is the content creation tools, currently in beta, which run the Writing Assistant workflow from your assistant. Those tools do write to your account: they create and change briefs, drafts, Content Profiles, and Knowledge Sources. They stop for your approval at each meaningful step, and they tell you when an action uses one of your organization’s allotted drafts.Nothing else in the MCP writes to Conductor, and no tool can delete anything.
They let you run Conductor’s Writing Assistant workflow in conversation with your AI assistant: build a brief, review it alongside a baseline content score, generate a draft, then refine it with title, meta description, and internal-link suggestions. See Content creation for the full workflow and Content for the tools themselves.The feature is in beta. Contact your Customer Success Manager, or email [email protected], and mention content creation in the MCP.
While content creation is in beta, those tools are available only through custom connections to Conductor’s MCP. Native connectors published in an LLM’s own marketplace run an older, reviewed tool set and won’t show them.This is the general pattern rather than a one-off: custom connections always carry the most up-to-date set of tools, while marketplace apps lag behind. If you need the newest capabilities, connect through a custom connection.
Yes. Creating a brief creates a draft, and generating content works against it, so content work through the MCP draws on the same allotment as work you do in Writing Assistant. The MCP tells you when an action uses a new draft, so you can keep an eye on usage before approving.
Do you already have an active, unfinished call to a content tool? Brief and content generation take time—and Conductor processes only one call to one these tools at a time. So, if you make more than one call to a tool before an existing call finishes, you’re initial call will error out. Wait until your first call finishes before making another.
Capabilities like pulling exact citation URLs for every prompt or integrating directly with third-party BI tools aren’t available yet. Several capabilities like these are on our roadmap.If there’s something you’d like the MCP to do, share it with your Conductor team as product feedback—it helps us prioritize.
Basic MCP queries like “show me my brand mention coverage for the topic of SEO” work across supported assistants, including Copilot. The richer, multi-step skills (audits, dashboard-style summaries with visualizations) are currently optimized for Claude and ChatGPT, where results are most reliable. We’re actively exploring how to bring the full skills experience to Copilot.One other difference: Claude and ChatGPT show you each tool call the assistant makes, while Copilot runs them behind the scenes. You can ask Copilot to report which tools it used if you need it.
No. Skills can check their own version against the latest published version, so you stay current without manually re-downloading.

​
Connecting to AI Platforms

To start, we allow connections to:
  • OpenAI’s ChatGPT Pro, Business, or Enterprise plans, using a plugin available in their plugin directory. Because ChatGPT and Codex share the same plugin install, one setup covers both.
  • Anthropic’s Claude Pro, Max, Team, and Enterprise plans, using either Conductor’s connector in Claude’s directory or a custom connector you can build. One setup covers Claude, Claude Cowork, and Claude Code.
  • Microsoft’s Copilot. Note that access can vary across plans and access level. Check with your IT team to verify what you have access to.
  • Perplexity’s individual Pro plans and Enterprise plans.
  • Most other AI surfaces that allow connections to MCP servers.
The connectors are available to all Conductor customers with access to supported LLMs. You will need an active Conductor login to authenticate the connection.
The most common reason you don’t see it has to do with permissions or security settings at your organization. Many apps need to be requested and approved by your IT team.How that approval works depends on the platform. In Claude, on Team and Enterprise plans an Owner or Primary Owner has to add Conductor to the organization before members can connect to it. In ChatGPT, a Workspace admin may need to allow the plugin on your account. In Copilot, you might have to click Request approval and wait for your IT team.If you can’t get the native app enabled, check whether your platform also supports a custom connector—in Claude, for example, you can add one yourself on Pro and Max plans. Your platform’s setup guide covers every path available to it.Also, be sure to note which LLMs Conductor supports with a native app in their ecosystem.
When Conductor releases new tools for your MCP connections to use, we add them to our MCP server, and you can then immediately update your connection to get access.New versions of native apps, plugins, and directory connectors—like Conductor in ChatGPT, Claude, or Copilot—go through an approval process when submitted to an LLM. Because of this, there can be a lag between when we submit a new MCP version with new tools and when those new tools appear for use in the LLM apps.This lag may be as little as a few days or as much as a few months.
Note that this only affects which tools are available—the underlying data is equally fresh either way. Whichever connection you use, you’re querying the same live Conductor data.
Whichever your platform supports. Both connect to the same Conductor MCP server, so the practical difference is setup effort versus how soon you get newly released tools.A native connector—Conductor’s listing in your AI platform’s directory or marketplace—is the faster setup. You find Conductor, click to connect, and authenticate. There’s no server URL or API token to paste. A custom connector takes a few more steps, but it always runs Conductor’s newest MCP version rather than waiting on the platform’s app-review cycle.Which paths are available depends on the platform:Custom connector steps for every platform—plus anything Conductor doesn’t document directly—live in Custom Connectors.
Every native app, plugin, and directory connector Conductor publishes goes through the AI platform’s own review process. When Conductor releases a new MCP tool, there can be a lag between our release and the platform approving the version that includes it.If a tool you expect is missing, start by making sure your connection is current—reinstall or reconnect it, and where the platform lists versions, choose the newest one. If it’s still missing and your platform also supports a custom connector, that path always runs Conductor’s newest MCP version.
Yes—as a custom connection. Conductor isn’t listed in Google’s app directory yet, so there’s no one-click connector to install. We’re working with the Google team on that process, similar to how Conductor is listed with ChatGPT and Copilot. In the meantime, both Gemini surfaces accept a custom MCP server:
  • Gemini Enterprise connects to Conductor as a Custom MCP Server data store in the Google Cloud console. See Connect Google Gemini Enterprise with Conductor for the full setup.
  • The Gemini app at gemini.google.com accepts custom apps, which you add by linking an MCP server URL under Settings & help > Connected Apps.
Both paths authenticate with OAuth and need client credentials issued by Conductor. Contact Conductor to request them.
Yes. Through Conductor’s partnership with Optimizely, Conductor is available in Optimizely Opal’s agent network, so Opal agents can work with your Conductor data while you work in Optimizely. The connection runs through Conductor’s MCP, so your organization needs an active Conductor subscription that includes MCP access.To set up Conductor in Opal, refer to this article in Optimizely’s Help Center or contact your Optimizely account team.
In addition to authenticating connections through OAuth, you can authenticate custom connections with an API key. Review the Custom Connectors article to learn more.
Generally speaking, yes! Conductor’s MCP server can connect to any AI tool or platform that supports MCP connectivity via OAuth or API bearer token. Setup for a specific tool may need some extra configuration details. Reach out to your Conductor team with the name of your tool, the authentication method it supports, and any setup documentation or screenshots, and we’ll help you work through the connection.

​
Subscription, Usage & Pricing

Every Conductor subscription includes allocated MCP tool calls, refreshed each contract year. If you’re an account admin, you can see the exact number included with your plan—and how much you’ve used—on the Subscription page in Conductor. If you’re not sure, your CSM or Account Manager can confirm your allocation.
Mcp Subscription
Your allocation refreshes automatically on an annual basis. You don’t need to purchase additional tool calls to keep the same allocation from year to year—your plan’s included calls renew for you.
Once you’ve used the tool calls included with your plan for the period, new requests are declined until more calls are added. You’ll never be charged for overages—requests simply pause rather than billing you unexpectedly. If you need more, reach out to your CSM or Account Manager, who can help you add additional tool calls to your subscription.
A tool call is counted only when the assistant successfully retrieves data from one of Conductor’s Data or Account Configuration tools. If a request fails or returns no data, it isn’t counted. A few things to know: a single question can trigger several tool calls (the assistant may look up your configuration, then pull the data, then drill in), and a skill can use several calls in one run. Only Conductor’s data tools draw down your allocation. Behind-the-scenes steps like the assistant loading the list of available tools, or a support skill reading Conductor’s help content, don’t count.
Calls to Content tools are not counted against your MCP tool call allotment. Instead, any calls to Content tools that create a Writing Assistant draft in the platform use a draft from your organization’s draft allotment.
Different queries and skills use different numbers of tool calls depending on how much data they retrieve. If you’re planning heavier usage—like running several skills (your own or from the Template Gallery)—your CSM or Account Manager can help you estimate the tool calls involved so you can plan your allocation with confidence.
The best estimate is based on your own usage. Your CSM or Account Manager can review your recent consumption, account for any one-off spikes, and help you project the tool calls you’re likely to need for the rest of your term—so you purchase the right amount rather than guessing.
Account admins can view usage on the Subscription screen in Conductor, at both the account level and the user level. In the Data section, click View by user to see each user’s tool-call count for the current subscription period—searchable by name or email—next to your plan’s allocation and how much of it you’ve used.
Conductor’s MCP is now available to all customers. To keep the service fast and reliable for everyone, we’ve introduced rate limits—30 requests per hour per user and 120 requests per minute across the system—so that no single user or agent can affect everyone else. Your plan also now includes a set allocation of MCP tool calls. Rest assured, you’ll never be charged for overages.
Each plan comes with a pre-allocated number of tool calls. If you use up your allocation, you can choose to add more—but you’ll never be charged for overages. Your CSM or Account Manager can share details on adding tool calls if you need them.
We review rate-limit increases on a case-by-case basis. If the current limits are getting in the way of your workflow, contact your Conductor team with details on what you’re trying to do, and we’ll assess whether an adjustment is possible.

​
Access & Troubleshooting

This usually means one of a few things: MCP access isn’t yet enabled on your subscription, your plan term has ended, or you’ve used up your tool-call allocation for the period. Start by confirming your Conductor plan includes MCP access, then reach out to your CSM or Account Manager—they can verify your entitlement and help restore access.
MCP access follows account membership in Conductor. If a specific user can’t connect while others can, the most common reason is that they haven’t been added to that Conductor account yet. Ask your admin to confirm the user is a member of the account they’re trying to reach.
The MCP returns data for the Conductor accounts you’re a member of. If you expected to see more accounts than are showing up, confirm with your admin that you’ve been added to all of those accounts—access to each account’s data depends on your membership.
A 500 error is most often a temporary service disruption. Wait a moment and try your request again. If it keeps happening, contact your Conductor team so we can look into it.

​
Security & Privacy

No. The Conductor connector is not used to train their models. You can query your enterprise data with confidence, knowing it remains secure and private to your organization.
We know security review is an important step. Our MCP Quick Start guide includes an example email you can copy and send to your security or admin team to kick off their review. If your team has deeper due-diligence questions, your Conductor team can coordinate the right details with our product and engineering teams.