> ## Documentation Index
> Fetch the complete documentation index at: https://dev.alpharun.com/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Server

> Connect AI assistants and agents directly to your Alpharun data

Alpharun provides a hosted [Model Context Protocol (MCP)](https://modelcontextprotocol.io) server that lets AI assistants and agents — like Claude, ChatGPT, and Cursor — work directly with your Alpharun data. Your assistant can search customer interactions, read transcripts, analyze playbook performance, and answer questions about your team, all without writing code or exporting data.

## Why use the MCP server?

* **No data pipeline to build or maintain.** Your assistant queries Alpharun live, so there's nothing to sync and results are never stale.
* **Rich filtering built in.** Interactions can be filtered by teammate, date range, outcome, playbook criterion scores, signals, and custom fields — server-side.
* **Semantic transcript search.** Search calls by meaning ("customers confused about pricing"), by keyword, or both — capabilities that would be expensive to replicate on top of a raw data export.
* **Analysis-ready context.** The server exposes your playbooks, criteria, outcomes, signals, and custom field definitions so assistants understand your data model, not just raw records.

## Setup

Follow the step-by-step setup guide in our help center: [Alpharun MCP](https://help.alpharun.com/en/articles/12880001).

## What your assistant can do

Once connected, an assistant has access to tools for:

| Area                  | Capabilities                                                                                                                                             |
| --------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Workspace context     | Discover your playbooks, custom fields, signals, and company configuration                                                                               |
| Customer interactions | Search, filter, and count interactions; drill into a single interaction with full criterion scores, commentary, extracted data, and complete transcripts |
| Transcript search     | Find calls by keyword, semantic meaning, or a hybrid of both                                                                                             |
| Playbook analytics    | Inspect criterion grading instructions, score distributions, outcome correlation, and section applicability                                              |
| Team performance      | Review per-teammate scores, trends, and comparisons to company medians                                                                                   |

## When to use MCP vs. REST API vs. webhooks

Each integration surface is built for a different job:

| You want to...                                                                                                            | Use            |
| ------------------------------------------------------------------------------------------------------------------------- | -------------- |
| Let an AI assistant or agent explore, analyze, or answer questions about your call data                                   | **MCP server** |
| Push data into Alpharun or automate workflows (create contacts, upload recordings, invite teammates, manage job openings) | **REST API**   |
| Read specific records into your own system as part of an integration                                                      | **REST API**   |
| React in real time when an interaction is analyzed or an interview completes                                              | **Webhooks**   |

A good rule of thumb: the REST API and webhooks are for *systems* talking to Alpharun; the MCP server is for *AI assistants and agents* talking to Alpharun.

<Note>
  If your goal is AI-powered analysis of your Alpharun data, connect through the MCP server rather than bulk-exporting interactions and transcripts via the REST API into your own datastore. A copied dataset goes stale immediately, needs its own search and filtering infrastructure, and loses the playbook context that makes the data meaningful — the MCP server gives your assistant all of that out of the box.
</Note>

## Questions?

Reach out to [support@alpharun.com](mailto:support@alpharun.com) — we're happy to help you get connected.