FAQs

Q: What is MCP Server?

A: MCP (Model Context Protocol) Server is the integration layer that connects Responsive to external AI tools, enabling those tools to securely access:

• Trusted Content Library knowledge

• Real-time project data

• Approved, governed responses

It allows users to stay in their AI tool while accessing Responsive intelligence.

Q: What problem does MCP solve?

A: It eliminates context switching and risk by:

• Bringing trusted knowledge into AI workflows

• Reducing hallucinations

• Maintaining governance and permissions

• Allowing users to work inside tools they already use

Q: What are the primary use cases for MCP Server?

A:  There are several:

1. Knowledge retrieval

• Ask questions against Content Library

• Generate grounded responses

• View sources

2. Project visibility

• View projects, status, deadlines

• See assigned questions

• Track progress

3. Response generation

• Generate answers to questions

• Use approved content

• Validate with TRACE + sources

Q: What should customers start with?

A: We recommend that you begin with simple use cases (search, summarization, and drafting) and then expand into project insights and response generation

Q: How should teams validate outputs?

A: We recommend that you:

• Always check sources

• Use TRACE Score

• Reinforce governance practices

Q: How do I know which tool I need, MCP Server or the Responsive App on ChatGPT?

A: There are a number of differences between the two:

• The MCP Server is the more robust, enterprise-grade integration. It provides a governed connection to Responsive data, enabling access to approved content, supporting context (like sources and TRACE Score), and real-time project information directly within AI tools.

• The Responsive App on ChatGPT, by contrast, is a marketplace-based experience designed for discoverability and ease of use within platforms. It offers a simplified version of functionality, focused on retrieving compliant answers from the Content Library, but does not include the full capabilities or configurability of MCP Server.

Q: Must the Responsive App appear in the AI platform’s marketplace to be available?

A: No, Responsive MCP server can function without having a corresponding marketplace app.

Q: Is the Responsive App on ChatGPT intended to replace Ask?

A: No. Responsive on ChatGPT (via the MCP Server) is not a replacement for Ask.

 It is an additional access point that allows users to leverage Responsive knowledge within external AI tools.

• Ask remains the native, fully governed SRM experience

• MCP extends that same knowledge into external workflows (ChatGPT, Copilot, Claude)

Using MCP Server embeds Responsive where users already work, rather than forcing workflow changes.

Q: Can I have both the MCP Server and the Responsive App on ChatGPT at the same time? How are they differentiated?

A: Yes. You can rename the MCP Server to something meaningful. The app cannot be renamed. We recommend using clear naming conventions to avoid confusion.

Q: Does the MCP Server respect permissions and security?

A: Yes. it enforces:

• User permissions

• Collection access

• Source-level permissions (e.g., SharePoint, Drive)

No unauthorized data is exposed.

Q: What are the prerequisites to set up the MCP Server?

A: There are different prerequisites for different users.

• For admins:

  • Admin access in ChatGPT/Copilot/Claude

  • Ability to configure connectors (MCP)

• For end users:

  • Valid Responsive license

  • Access to content/projects in Responsive

• For the organization:

  • Agentic pricing (Enterprise or add-on for other tiers)

Q: What are the key set up steps?

A: In general, you will perform the following steps:

1. Add the MCP Server URL

2. Configure authentication (OAuth)

3. Log into Responsive

4. Select your instance

5. Start querying data

For details, refer to Using MCP Server with Generative AI Tools.

Q: Can users use the MCP Server without a Responsive license?

A: No.  A valid Responsive license is required.

Q: Can the MCP Server be enabled for specific users only?

A: Not currently.

• It is controlled at org level today

• Org-level controls are being built

• No user-level granularity is offered yet

Q: Which platforms are supported today?

A: Responsive supports the following platforms:

• ChatGPT (fully supported)

• Claude (supported via MCP Server, not the marketplace app)

• Copilot (supported, but not fully GA)

Q: Can MCP Server connect to other LLMs beyond ChatGPT, Claude, Copilot?

A: Yes. MCP Server is designed to work with any LLM that supports the MCP protocol, though only these three are officially validated today.

Q: What are the differences between the the MCP Server and the Responsive App on ChatGPT?

A: There are several:

Q: Why do we recommend the MCP server over the Responsive App on ChatGPT?

A: The app is limited to search only:

• No project workflows

• No TRACE Score

• No deeper context

Q: Will the Responsive App on ChatGPT improve over time?

A: Yes. Our goal is to bring more MCP Server capabilities into the app, but the app is controlled by platform providers (OpenAI, Microsoft, Anthropic), and there are no guarantees that all features will be allowed.

Q: Can users see sources and validate answers?

A: Yes. Users can:

• View sources

• Click into original content

• Validate accuracy

Q: Does the MCP Server support TRACE Score?

A: Yes for generated responses, but it doesn’t always appear for a simple content retrieval unless explicitly generated.

Q: Can users complete work directly from ChatGPT?

A: Currently, you cannot:

• Write back to projects

• Update project status

• Save answers to library

Q: What data does the MCP Server access?

A: Data is retrieved from:

• Content Library (Q&A + documents)

• Project data

• External sources (SharePoint, Google Drive)

Q: Does the MCP support external websites?

A: Not currently. Only configured external sources are supported.

Q: Can users control which sources the MCP Server uses?

A: Content filtering (like Ask has) is not yet supported, but it is coming.

Q: Which instance does the MCP Server use if a user has more than one?

A: Users will select an instance during login, and MCP Server will use that instance.

Q: Can users switch instances easily?

A: Users must disconnect and reconnect; they cannot dynamically switch mid-session.

Q: Can users differentiate between multiple MCP Server connections?

A: Yes, by connector name. Each connector must have a unique name.

Q: Does MCP Server usage flow back into Responsive analytics?

A: Limited. Currently, responses are treated as manual answers.

Q: Are there token limits?

A: Responsive has no token limits today, subject to reasonable use policies. The LLM providers do (based on their contracts).

Q: Are there usage limits?

A: Responsive currently supports approximately ~100 API requests per second. Usage is subject to reasonable use policies.

Q: Is customer data used to train LLMs?

A: No. Data is retrieved from Responsive, and it is not used for model training.

Q: How does Responsive reduce hallucinations?

A: By grounding responses in:

• Approved Content Library

• Verified sources

• TRACE Score validation

Q: Can I enable the MCP Server for specific users?

A: Currently MCP server access is controlled at the organization level. You cannot limit access to specific users at this time.

Q: Who can configure the MCP Server?

A: This is determined by the AI platform (ChatGPT, Claude, Copilot). Typically, only users with admin permissions for those tools can configure them.

Q: Does the MCP Server configuration apply to one user or the entire workspace?

A: The configurations are applied at the workspace level. After being configured, it is available to all the users in that workspace.

Q: What is the release cadence for the MCP server vs. the Responsive App on ChatGPT?

A: The cadence for the Responsive app depends on external platform approval cycles (OpenAI, Anthropic, Microsoft). The cadence for MCP Server follows Responsive’s standard release cadence, with the flexibility to release features faster when ready.

Q: Is the Responsive App on ChatGPT connected to a customer’s specific instance?

A: Yes. Users authenticate with their Responsive credentials, and access is tied to their instance.  However, the experience is limited compared to MCP Server.

Q: How should customers differentiate between the MCP Server and the Responsive App on ChatGPT experiences?

A:There are a number of differences between them:

• The MCP Server can be custom named/configured (recommended for clarity)

• The app is fixed and marketplace-based

• The MCP Server provides deeper, contextual, governed intelligence

Q: What data sources can the MCP Server access?

A: It can access:

• Content Library (Q&A pairs, documents)

• Real-time project data

• External sources (e.g., SharePoint, Google Drive) if configured in Responsive

Access is governed by existing permissions and configurations.

Q: Can the MCP Server analyze content (e.g., deduplication, semantic analysis)?

A: No. It currently does not support:

• Deduplication

• Identifying semantic duplicates

• Advanced content analysis

Q: Can the MCP Server summarize tasks or retrieve project-related insights?

A: Yes. It can retrieve project questions and data for users and support summarization use cases.

Q: Can the MCP Server generate reports or analytics (e.g., what are the most asked questions)?

A: Not currently. Reporting and analytics capabilities are not supported today.

Q: Can the MCP Server create graphs, slides, or deliverables using Responsive data?

A: It depends on the AI tool (ChatGPT, Claude, etc.). Responsive does not provide structured data (e.g., usage metrics) for visualization at this time.

Q: Does TRACE Score appear in ChatGPT via MCP?

A: TRACE Score and supporting context are available through the MCP Server (but not in the Responsive App for ChatGPT).

Q: How does the MCP Server determine writing style?

A: The writing style is determined by the external LLM (ChatGPT, Claude, etc.), not Responsive.

 Responsive provides the grounded data; the LLM generates the output.

Q: Which Responsive instance does the LLM connect to if a user has more than one?

A: It connects to the instance the user is currently logged into. Users cannot easily switch instances within ChatGPT, they must disconnect and reconnect to select a different instance.