Agentic AI: The EdgeTier MCP server

What is MCP?

The Model Context Protocol (MCP) is an open standard that gives AI assistants a simple, unified way to access and interact with external data and tools.

Without an MCP server, connecting an AI assistant (e.g. Claude) to a platform like EdgeTier would require custom, one-off API integrations and someone technical to build and maintain them. Our MCP server provides a single, standardised connection that exposes all available capabilities at once, updates automatically as new features are added, and can be configured with a few clicks rather than lines of code.

---

API vs. MCP — what’s the difference?

If you’ve connected tools via APIs before, here’s how using an MCP server compares:

Integrating with APIs

• Ideal for software-to-software integrations
• Configured via coding (e.g. JSON)
• Individual capabilities may require separate integrations or configurations
• Requires maintenance when capabilities are added or changed

Integrating with MCP

• Ideal for software-to-AI integrations
• Configured via point-and-click interfaces with a simple link
• A single integration exposes all available capabilities on demand
• New and updated capabilities are automatically included

Unlike APIs, which need to be set up once per capability and revisited whenever something changes, MCP integrations only need to be set up once. From that point on, AI assistants can access new and existing capabilities automatically, without any additional configuration. This reduces the time and resources spent on custom integrations and helps AI agents make smarter, faster decisions without losing context.

---

How to connect

EdgeTier’s MCP server exposes your contact centre intelligence to any compatible AI assistant. Once connected, the AI can reach directly into your EdgeTier data, pulling metrics, reading conversations, detecting anomalies, and running qualitative analysis, all through natural language, without leaving the chat interface.

Like the other features in the EdgeTier platform, access to our MCP server is managed with roles and permissions. For security reasons, before you connect, you’ll need to get in touch with your EdgeTier CSM and ask to turn on MCP server access for your organisation. Then, you can decide which team members have access through Settings > Roles & Permissions.

Once enabled, users with permissions will have secure, scoped access to your organisation’s data. They’ll only be able to access the data and tools they can already see in the platform.

You can find more detailed information on connecting to our MCP server in our documentation, here!

---

Capabilities overview

Your MCP server can only access data you already have the permission to view and manage within EdgeTier.

Once connected, your AI assistant can access the following EdgeTier capabilities:

Metrics & Volume

Tool	What it does
get_interaction_aggregates	The primary metrics tool. Returns volume, AHT, CSAT, NPS, Experience Score, and frustration rate — grouped by contact reason, agent, channel, or time period.
get_anomalies_by_time	Returns statistically significant spikes or drops detected during a given window — useful for identifying what changed and when.
get_ongoing_anomalies	Shows currently active, unresolved anomalies for live situation awareness.

Conversation Intelligence

Tool	What it does
get_themes_from_interaction_groups	AI-powered theme analysis across a set of interactions. Identifies dominant topics, what proportion of conversations relate to each, and provides examples. Essential for understanding why metrics are moving.
search_interactions	Finds specific conversations by emotion, CSAT score, agent, tag, message count, or free-text search.
get_interaction_messages	Returns the full message-by-message transcript of a specific conversation.

Quality Assurance

Tool	What it does
get_agent_scorecard_pass_rates	Aggregated scorecard pass rates per agent, including evaluation count, coverage %, and mean score. Sorted lowest-first by default.
get_scorecard_evaluations	Paginated list of individual evaluations, filterable by agent, scorecard, pass/fail status, and date range.
get_scorecard_evaluation_details	Full question-by-question breakdown of a single evaluation, including reviewer comments.

Tags, Agents & Search

Tool	What it does
get_tag_groups / get_tags	Browse the full taxonomy of contact reasons, queues, channels, and AI-assigned categories in your EdgeTier instance.
search_agents	Find agent IDs by name or email for use as filters in other queries.
create_text_search_params	Filter any query by what was actually said in conversations — competitor mentions, product names, promotions, or any keyword.

---

Example prompts

Once connected, describe what you need in natural language. Here are some prompts that work well with the EdgeTier MCP server:

Operations overview

“We had an outage on Tuesday. Pull all EdgeTier contacts referencing the issue, group them by customer tier from Salesforce, and tell me the total revenue exposure of affected accounts”

Agent performance

“It’s end of month. Pull all EdgeTier QA scores for the last 30 days, identify any agents sitting below the team average, then check their contact volume and average handling time to see if it’s a capacity issue or a quality issue. For the bottom 3, draft a Slack message to their team lead with a summary of where they’re falling short by scorecard category, and suggest 2-3 coaching focus areas based on the pattern.”

Topic deep-dive

“Pull all EdgeTier contacts from the last 7 days mentioning ‘billing error’, cross-reference with Salesforce to find which are enterprise accounts, and tell me the total ARR at risk.”

Anomaly investigation

“Every morning, pull the overnight EdgeTier anomalies and post a summary to the #cx-alerts Slack channel so the team sees it before standup”

Quality audit

“Find all customers who contacted support more than 3 times in the last month and match them against Salesforce — are any of them flagged as churn risk? Cross reference with their agent in EdgeTier Coach”

---

Tips and Best Practices

EdgeTier tells you what’s really happening with your customers, now also in the context of everything else you know about them.

Be specific about time periods

The AI works best when you specify “yesterday”, “last week”, or a concrete date range. Without a time period, it will use a sensible default for the question you’ve asked.

Drill down with follow-up questions

The MCP connection is stateful within a session, meaning you can ask multiple follow up questions in the same chat. Start broad (“give me an ops report”) then drill into specifics (“now look at the billing contact reason and show me examples from frustrated customers”).

Combine with other MCP connectors

MCP supports multiple simultaneous connections. If your assistant is also connected to tools like Slack, Shopify, Linear, or Notion, you can push EdgeTier insights directly into your workflows, for example: “Create a linear ticket for this anomaly” or “Send a summary of this to the #ops Slack channel.”

Filter by emotion for targeted analysis

EdgeTier automatically detects customer emotions — frustration, praise, gratitude, empathy — across all interactions. Use these as filters: “Show me the top contact reasons among frustrated customers this week.”

There is so much more you can do with this connector, so let’s talk about your use case! If you have any questions about this release, please reach out to Customer Success.

FAQs

• How do I get access?
  • You will need to request access from your Customer Success Manager. Once access is enabled, ensure the ‘Use MCP’ permission is enabled for your role.

• How is it different to ‘Ask Spotlight’? Which should I use?
  • Ask Spotlight is EdgeTier’s built-in AI assistant, the easiest way to ask natural language questions about your contact centre data, right inside the platform. No setup needed. For most users, this is the right place to start.
  • The EdgeTier MCP is for technical teams who want to bring EdgeTier data into external AI tools (like Claude or custom internal copilots) as part of a broader workflow. It requires configuration and is best suited to developers building cross-platform AI integrations.
  • Not sure which to use? If you’re an EdgeTier user looking for faster insights, Ask Spotlight has you covered. The MCP is there when your team needs to go further.

• How does my data work with the EdgeTier MCP server?
  • Data retrieved via the MCP is used by the external AI tool to respond to your query. In short, they use your EdgeTier data. Please refer to your AI provider’s data policy for details on how that data is handled.