# MiiTel MCP

> MiiTel analyzes phone calls with your AI agent, scoring call quality while identifying sentiment and providing detailed coaching insights for sales teams. It gives you full text transcripts, tracks talk ratio and speech rate, and monitors overall team performance metrics. You connect it to any MCP-compatible client and start reviewing calls instantly.

## Overview
- **Category:** customer-support
- **Price:** Free
- **Tags:** call-transcription, speech-analytics, sentiment-analysis, sales-coaching, call-recording, ip-telephony

## Description

This server connects your AI agent directly to MiiTel's call analytics engine, letting you review sales calls and team performance data using natural language prompts. You won't just get transcripts; you'll get full coaching insights into how the conversation went down.

**Getting Call Data and Context**

You can start by pulling a list of all your past interactions with `list_calls`, which provides an exhaustive record of every call history in your MiiTel account. Need to know who was on the line? Use `get_call_details` to pull specific performance metrics and metadata for any single call ID you care about. If you're tracking meetings instead, `list_online_meetings` generates a list of past online sessions, giving you the dates and primary participants involved in those calls.

To keep your analysis grounded, you can cross-reference everything. You've got `list_synced_contacts`, which fetches an up-to-date directory of contacts synced straight from your CRM system. If a call relates to an online meeting, `get_meeting_details` pulls the full recorded data, listing participants and noting the final outcome.

For internal management, you can pull a list of all active user accounts within MiiTel by calling `list_miitel_users`, so your agent always knows who's on the books. You won't need to manually search for names or dates; these tools get it ready for your AI client.

**Deep Call Analysis Mechanics**

The server handles more than just audio dumps. When you analyze a call, your agent pulls full text transcriptions of sales or support calls and gives you instant summaries detailing the main topics discussed—like spotting competitor names that came up during the chat. The system calculates key speech metrics like talk ratio, average speech rate, and even flags overlaps or excessive silence throughout the recorded conversation.

Beyond the mechanics, it analyzes emotional tone shifts. It tracks sentiment and energy levels across the entire call duration, flagging specific moments when the customer's mood suddenly changes or improves. You get total text transcriptions alongside instant summaries that help you pinpoint those critical turning points in a sale.

When tracking team performance, the system pulls aggregate metrics for reporting. You can monitor total call volume, average conversation length, and identify top performers across your sales force, giving you clear data on what's working well and where coaching is needed. Your agent combines all this—the raw list of calls from `list_calls` with the deep performance data from `get_call_details`, cross-referenced against contacts via `list_synced_contacts`—to give you a complete picture every single time.

## Tools

### get_call_details
Retrieves specific performance and metadata details for one single call ID.

### get_meeting_details
Gets the recorded details, participants, and outcome data from a specific online meeting.

### list_calls
Provides an exhaustive list of all call history records within your MiiTel account.

### list_synced_contacts
Fetches an up-to-date list of contacts synced from your CRM system.

### list_online_meetings
Generates a list of past online meetings, including their dates and primary participants.

### list_miitel_users
Pulls a directory listing of all active user accounts within the MiiTel system.

## Prompt Examples

**Prompt:** 
```
Show speech analytics for Mike's latest sales call.
```

**Response:** 
```
Call ID: call_890 (Mike vs Acme Corp). Duration: 14m 20s. Talk Ratio: 45% Mike / 55% Customer (Excellent ✅). Speech Rate: 160 WPM. Interruptions: 2. Longest silence: 4.5s. Emotion: Positive trend at the end.
```

**Prompt:** 
```
Get the transcript summary for call call_890.
```

**Response:** 
```
Transcript (call_890): Mike asked discovery questions about Acme's current CRM setup. Customer mentioned 'Salesforce is too expensive'. Mike pitched the Pro Plan at $45/mo. Customer agreed to a follow-up demo next Tuesday. Competitor mentioned: Salesforce.
```

**Prompt:** 
```
Show today's call volume and team performance.
```

**Response:** 
```
Team Performance (Today): Total calls: 145. Total connected: 42 (28%). Avg duration: 4m 12s. Top performer: Sarah (18 connected, avg talk ratio 48%). Lowest talk ratio: John (85% talking - needs coaching ⚠️).
```

## Capabilities

### Analyze call mechanics
It calculates key speech metrics like talk ratio, average speech rate, and identifies overlaps or excessive silence during recorded conversations.

### Extract transcripts and summaries
You pull full text transcriptions from calls and get instant summaries of the main topics discussed (e.g., competitor names mentioned).

### Track emotional tone changes
The system analyzes sentiment and energy levels throughout a call, highlighting moments when the customer's mood shifts or improves.

### Monitor team performance data
It pulls aggregate metrics like total call volume, average duration, and top performers for easy reporting.

### Retrieve related context
You can cross-reference the analyzed call against existing CRM contacts or recorded online meeting histories using tools like `list_synced_contacts`.

## Use Cases

### The Sales Manager needs coaching on discovery calls
A manager wants to review Mike's pitch. Instead of re-listening, they prompt their agent: 'Show speech analytics for call_890.' The agent runs `get_call_details` and replies with the talk ratio (e.g., 45% vs 55%), showing exactly if Mike talked too much or not enough.

### QA needs to audit compliance across a team
The QA specialist asks their agent to 'List all calls from last week.' The agent uses `list_calls` and filters the results, providing a list of IDs. The specialist then runs `get_call_details` on each ID to check for required legal disclosures.

### Customer Success needs to spot churn risk
A CS lead wants to know why Account X is struggling. They prompt: 'Check sentiment history for Acme Corp.' The agent analyzes the last five calls, noting a consistent negative trend in emotion analysis, alerting them before the client complains.

### The team needs context on a new meeting
A rep finishes an online call and asks their agent to 'Get details for this meeting.' The agent runs `get_meeting_details` and provides not only the recording link but also who attended, which helps the rep follow up with the right people.

## Benefits

- **Stop guessing about performance.** With `list_calls` and `get_call_details`, you get a full history of every interaction, allowing you to audit patterns that individual notes miss.
- **See exactly where the conversation went off track.** The AI tracks sentiment changes moment by moment. You'll see if the customer got frustrated after a certain pitch point, giving reps immediate feedback on tone.
- **Cut through the noise with transcripts.** Instead of reading pages of dictated notes, you ask your agent to 'Summarize call_890.' It pulls out key agreements and competitor names in seconds.
- **Benchmark entire teams instantly.** Use performance analytics to track total calls connected versus average duration. You can immediately spot which reps need volume coaching and who needs talk ratio help.
- **Connect the dots automatically.** By cross-referencing call data with `list_synced_contacts`, your agent doesn't just analyze a conversation; it ties that analysis back to the customer’s entire history in your CRM.

## How It Works

The bottom line is that MiiTel takes recorded calls and turns them into actionable performance reports for your AI client.

1. First, you subscribe to the MiiTel server and provide your unique Access Token.
2. Next, your AI client runs a prompt against the tools (e.g., 'Show me Mike's last call metrics').
3. Finally, the agent receives structured data—like talk ratio or sentiment trends—and presents it back to you in plain text.

## Frequently Asked Questions

**How do I use list_calls to see all my calls?**
You prompt your agent with 'List MiiTel call history.' The agent executes `list_calls` and returns an exhaustive record of every interaction, giving you the IDs needed for further analysis.

**Can I get coaching insights using get_call_details?**
Yes. After getting the details from a specific call ID, your agent can analyze the talk ratio, overlaps, and speech rate to pinpoint exactly where the rep needs improvement for better coaching.

**What is the difference between list_calls and list_online_meetings?**
They track different types of interactions. `list_calls` handles voice/telephony records, while `list_online_meetings` tracks structured virtual meeting history.

**Can I use MiiTel MCP Server to find contact info?**
Yes. You can run `list_synced_contacts` to pull the latest data from your CRM, ensuring that any call analysis you perform is immediately associated with accurate customer records.

**What permissions are required when calling `list_miitel_users`?**
You need read scope access to list users. The API requires an authentication token with 'user:read' permission enabled. This ensures your agent can retrieve user accounts without needing write or modify rights.

**Does `get_call_details` provide the full conversation transcript?**
No, it primarily returns call metadata and links to the recordings. To get the actual text transcript, you must use the provided transcript URL found within the detailed response object.

**How do I filter results from `list_calls` by a specific date range?**
You pass start and end dates as query parameters in your request body. For example, set 'start_date' and 'end_date' to restrict the call history to the desired timeframe.

**Are there rate limits when I use `get_meeting_details` frequently?**
Yes, API requests are subject to standard rate limiting policies. For high-volume processing, implement a delay or consider batching your calls rather than making them sequentially.

**Can I access full call transcripts?**
Yes. Retrieve full text transcripts of calls, organized by speaker turns and timestamps.

**How does MiiTel authentication work?**
MiiTel uses an **Access Token** (Bearer auth) against `revcomm.miitel.jp/api/v2`.

**Does it analyze speech patterns?**
Yes. The API provides metrics like Talk/Listen ratio, interruptions, silences, and speaking speed (WPM).