> For the complete documentation index, see [llms.txt](https://docs.gradient.works/kb/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.gradient.works/kb/mcp/overview.md).

# MCP Overview

Work with Gradient Works in your AI tool of choice (Claude, ChatGPT, etc.) using natural language. The Gradient Works MCP arms your AI with the tools to speed up key RevOps workflows across territory planning & book of business management.

## What is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard that lets AI tools securely connect to the apps you already use and take action in them. Because it's a shared standard, any MCP-compatible AI tool can work with any MCP-enabled app, without a custom one-off integration for each pairing.

The **Gradient Works MCP** is that connection for your Gradient Works account. Once connected, your AI tool can read your routing, books of business, market map, and rep coverage data (and, with the right access, make changes), all from a natural-language conversation. The MCP server URL is:

```
https://agents.gradient.works/mcp
```

## What you can do with the Gradient Works MCP

Connect Gradient Works to your AI tool and ask it to do the work in plain language. A few examples:

| Use case                                        | Example prompt                                                                                                               |
| ----------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------- |
| **Investigate & debug routing automations**     | *"Which of my assignment flows route inbound leads, and where could a lead get misassigned or dropped?"*                     |
| **Understand & document your assignment flows** | *"Which of my flows route inbound leads, and how do they work? Trace the full path from form submission to rep assignment."* |
| **Prioritize & segment accounts**               | *"Score and segment our mid-market US-East accounts and tell me which segments are most promising by fit."*                  |
| **Analyze rep coverage**                        | *"Show me account coverage and activity by rep over the last month, and flag anyone underwater."*                            |
| **Create & manage dynamic books of business**   | *"Our SDR Jordan was just promoted to AE. Add them to the Enterprise target book and remove them from the SDR book."*        |
| **Plan & compare territory scenarios**          | *"Compare two territory scenarios in Carve and show me which reps gain or lose accounts."*                                   |

See the [full prompt library](/kb/mcp/prompt-library.md) for more ways to put your AI tool to work using the MCP.

## Getting started

* [API Keys & Connections](/kb/mcp/api-keys.md)
* [Setup: Claude (Desktop, Web, or CLI)](/kb/mcp/setup_claude.md)
* [Setup: ChatGPT or Codex](/kb/mcp/setup_chatgpt.md)
* [Flow Tools](/kb/mcp/flow-tools.md): create and edit Salesforce assignment flows that use Gradient Works actions

## Prefer programmatic access?

The same capabilities are available over a REST API for scripts, integrations, and automated workflows. Full reference documentation is at [agents.gradient.works/api/docs](https://agents.gradient.works/api/docs/).

## Access levels

Both API keys and MCP connections use the same two access levels:

|                        | Read-only | Full access |
| ---------------------- | --------- | ----------- |
| View and query data    | ✓         | ✓           |
| Create and update data | —         | ✓           |

**Read-only** is appropriate for most AI assistant connections. It lets your AI see your data and answer questions without the ability to make changes.

**Full access** allows creating and updating records. Use this when you need to automate writes, such as a script that runs distributions or creates market maps.

How the access level is assigned depends on the connection type:

* **API keys:** you choose the access level when you create the key.
* **MCP connections:** the access level you can grant is capped by your user role. **Admins can choose read-only or full access; all other users can connect read-only only.** You can't grant an MCP connection more access than your own role allows.

Admins can revoke an API key or disconnect an MCP connection at any time, and it takes effect immediately. See [API Keys & Connections](/kb/mcp/api-keys.md) for step-by-step instructions.