Get started with Front's MCP server [beta]

⚠️ Beta note: To get access to this feature, please join our community group and follow the instructions in the welcome post.

Front's MCP server lets you and your teammates connect Front to AI assistants like Claude, ChatGPT, Gemini, and Microsoft Copilot. Once connected, the AI assistant can use your conversations, draft replies, leave internal comments, and take other actions on your behalf — all while respecting the user permissions set within Front.

Because Front handles every channel — email, SMS, chat, and more — in one place, the AI assistant gets the full context of a conversation: messages across all channels, internal comments, tag and assignment history, and rule actions, all in one timeline.

MCP (Model Context Protocol) is an open standard that lets AI assistants connect to external tools. The MCP server specifically enables external AI assistants and similar tools to access a predetermined set of Front data and actions. Connecting to the Front MCP server means the AI assistant can do things like search your inbox, read a thread, draft a reply, leave an internal comment, or tag a conversation — as you, with your permissions.

Think of it as giving the AI assistant a key to Front that only works while you're authorized, and only does the things your role already allows.

Here are some common ways to put the MCP server to work:

Triage your inbox

"Look at my Open inbox and help me prioritize what to respond to first. Flag anything that mentions billing or churn risk."

Draft replies with broader context

"Find the most recent conversation with [customer]. Cross-reference their account in Salesforce, then draft a reply that addresses their question and acknowledges where they are in the renewal cycle."

Investigate a bug report

"Open the most recent conversation tagged 'bug.' Summarize the issue, search GitHub for related PRs, and post an internal comment for engineering with what you find."

Run a high-risk review

"Pull all conversations tagged 'high-risk' from the last 48 hours. Add an internal comment on each with a recommended next step, then give me a summary in chat."

Extract themes across many conversations

"Look at the 50 most recent conversations in the Support inbox. Group them by topic and tell me which themes are growing fastest week-over-week."

For more examples and to see what others are building, head over to the MCP space in the Front Community.

⚠️ Beta note: To get access to this feature, please join our community group and follow the instructions in the welcome post.

There are two ways to connect, depending on what your workspace and your AI client support.

The simplest path is finding Front in the connector directory of your AI assistant of choice and connecting in a few clicks.

If your AI assistant is managed by your company — as it often is in larger organizations — your IT or business systems team will need to allow the Front connector before it appears in your account. If you don't see it listed, that's the team to ask.

Available soon:

• Claude — Front will be available in the Claude connector directory at Customize > Connectors > Browse Connectors.

We're prioritizing directory listings based on customer demand. If your team relies on an AI assistant we haven't listed, let us know in the Front Community — that's how we decide what to add next.

If your AI assistant's directory doesn't have Front yet, you can still connect today using the custom path below.

If you want to use Front in an MCP client that doesn't (yet) have an official directory listing — or if you're building your own internal agent — your workspace admin can create a private developer app in Front and use its credentials to connect.

This works in any MCP-compatible client, including:

• Claude Code — Anthropic's command-line coding agent

• Cursor — the AI-native code editor

• ChatGPT's developer mode — for custom MCP server configuration

• Any other MCP-compatible client, including in-house harnesses built on Claude, OpenAI, or other SDKs

Setup involves creating the app in Front, configuring OAuth scopes, and pointing your MCP client at mcp.frontapp.com/mcp. Full instructions live in the Front Developer Portal.

For both connection options, Front admins can limit which read/write/send permissions the MCP server can use. These selected permissions define the full set (superset) of data and actions the MCP server can access. From there, each connected teammate is still restricted to what their own Front permissions allow. For example, your company might enable MCP server access for all Front inboxes, but because each teammate must sign in to their own Front account, the MCP server for that teammate will only be able to access the inboxes they can access in Front.

• Option 1: Official connector in an AI assistant directory — the permissions are controlled by your company's AI assistant administrator (typically the IT or business systems team managing your enterprise account, e.g. Claude Enterprise) depending on the controls the AI assistant supplies. When admins allow managed employees to connect Front in the AI assistant, they'll see a list of tools and permissions available through the connector and can configure which ones are enabled for their users, if such a permission model is supplied. The options available will vary by AI assistant.

• Option 2: Custom connector — the permissions are controlled by the Front admin who creates the developer app in Front, as described in our Developer Portal. When configuring the app in Front, the admin restricts the read/write/send permissions the MCP server can use, and that becomes the superset for any AI assistant connecting through it.

In both cases, the MCP server for that user is still restricted to what their Front role allows.

The MCP server currently lets you:

• Search and read conversations across every channel (email, SMS, chat, custom)

• Draft and send replies on the conversation's native channel

• Read and post internal comments

• Tag, assign, and change conversation status

• Search and update contacts

• See teammates, inboxes, and teams

More capabilities are in the works. If there's something you need that isn't supported yet, share your use case in the community — that's how we prioritize what ships next.

We're building this with active feedback from beta users, and we'd love yours. The MCP space in the Front Community is the place to:

• See what others are building and share your own workflows

• Ask questions and get help connecting

• Submit feedback and feature requests