Table of contents
Is there an official Intercom MCP server? Yes - Intercom runs a remote MCP server at mcp.intercom.com, giving MCP-compatible AI clients secure, permission-aware access to conversations, contacts, companies and Help Center articles. It’s maintained by Fin, the company that used to be called Intercom, Inc.
This information is correct as of September 2026.
Update, September 2026: Intercom’s parent company renamed itself Fin on 12 May 2026, and on 15 June 2026 agreed to a roughly $3.6 billion acquisition by Salesforce, expected to close in Q4 of Salesforce’s fiscal 2027. Neither change has touched the product name or the MCP endpoint: the help desk is still called Intercom, and
mcp.intercom.comis still the connection marketers and support teams use.
We cover an MCP server for one part of the martech stack most weeks on this blog, and Intercom (or Fin, depending which part of the business you mean) is a widely used one for customer messaging and support. If you’ve searched “Intercom MCP server” wanting a straight answer on whether it exists and how it’s different from the third-party options floating around, here it is.
What the Intercom MCP server actually does
Intercom hosts the server itself, so there’s nothing to deploy or run on your side. Once connected, an MCP-compatible AI client can:
- Search conversations and contacts using a query syntax (for example, filtering by state, source type or a text match inside a message)
- Retrieve detailed records for specific conversations, contacts or companies, including custom attributes and activity timestamps
- Read and update Help Center articles, so an agent can draft or revise support content directly rather than through a separate CMS step
- Maintain pagination across large result sets, so an agent working through thousands of conversations doesn’t need you to script the loop yourself
Authentication works two ways: an OAuth flow (recommended, browser-based, and the default for most MCP clients) or a direct Bearer token for simpler setups. Either way, the server inherits whatever permissions the connecting account already has in your workspace rather than granting a separate, broader service identity.
“This release is the clearest signal yet that MCP is becoming real production-grade infrastructure.”
David Soria Parra, Member of Technical Staff and co-inventor of MCP
Intercom MCP server options compared
The official server isn’t the only way to connect Intercom to an AI client. A community project, raoulbia-ai/mcp-server-for-intercom, takes a different approach built for speed over large conversation volumes.
| Official Intercom/Fin server | raoulbia-ai/mcp-server-for-intercom (community) | |
|---|---|---|
| Status | Actively maintained by Fin | Actively maintained, open source |
| Hosting | Remote, hosted at mcp.intercom.com | Self-hosted |
| Auth | OAuth 2.0 (recommended) or Bearer token | Your own Intercom API token |
| Permissions | Inherits the connecting account’s existing Intercom permissions | Configurable, since you run and scope it yourself |
| Region support | US and EU hosted workspaces (AU not yet supported) | Anywhere the Intercom API is reachable |
| Best for | Most teams wanting a supported, zero-maintenance connection | High-volume conversation analysis, or workspaces outside Fin’s supported regions |
For nearly every marketing team, the official server is the sensible starting point: no infrastructure to run, and it comes with Intercom’s own permission model already built in. See our full MCP servers list for how this compares to the analytics, CMS and workflow servers most teams connect first, or MCP vs API if you’re still deciding whether MCP is the right layer at all.
Practical marketing use cases
Support conversations are one of the richest, least-used sources of first-party customer language most marketing teams have. Once connected, an agent can turn that archive into working material:
- Sourcing real customer quotes: “Find positive conversations mentioning our new reporting feature, and pull the best quotes into my campaign brief”
- Building a feature request backlog: aggregating recurring asks across weeks of conversations into a ranked list for product and content planning
- Surfacing objections for sales enablement: what customers actually say when they hesitate, in their own words, not a paraphrased summary
- Connecting support themes to lifecycle messaging: pairing what’s coming up in conversations with where a customer sits in the customer lifecycle, so lifecycle campaigns respond to real friction instead of a generic send schedule
Security considerations
Support conversations are also where a lot of a company’s messiest first-party data lives: emails, order details, complaints, sometimes payment or account information. A couple of things are worth knowing before you connect any AI client at scale.
First, the official server’s permission model matters, but it isn’t a substitute for scoping the connecting account deliberately. If that account can see everything in your workspace, so can the AI client.
Second, the underlying protocol carries a general caveat that applies here as much as anywhere else.
“It was designed for interoperability, not containment.”
Nancy Wang, SVP and Head of Engineering and AI, 1Password, on the Model Context Protocol
In practice: prefer read-only scopes for anything that doesn’t need to write back to Help Center articles, review what the OAuth grant actually covers before approving it, and treat conversation data pulled into an AI conversation with the same care you’d apply to an export.
Getting started checklist
- Confirm your workspace’s hosted region. US and EU are supported; AU is not yet.
- Choose OAuth (recommended, browser-based) or a Bearer token for authentication.
- Add the MCP connection in your client. Claude, Claude Code, ChatGPT, Cursor, Windsurf and VS Code all have documented setup steps for
https://mcp.intercom.com/mcp. - Check the scopes the connecting account will grant match only what you need, particularly if any workflow will create or update Help Center content.
- Test with a narrow, read-only query first, before enabling anything that writes back to your workspace.
- If you’re outside the supported regions or need to process very high conversation volumes locally, evaluate the community server instead.
“These agents will be sellers and advisers, teachers and experts.”
Eoghan McCabe, CEO of Fin (formerly Intercom)
About Growth Method
An MCP server like this one is only valuable if what it surfaces actually changes what you do next. Growth Method is the agentic marketing platform for B2B teams: plan, launch, and analyse campaigns with pre-built AI agents that work from your live marketing data. Connect Intercom alongside your analytics and CMS tools, and both your team and its agents can turn a pattern spotted in support conversations into a tracked campaign, planned against a real baseline and read against a real result, not just a quote pasted into a brief.
We are on-track to deliver a 43% increase in inbound leads this year. There is no doubt the adoption of Growth Method is the primary driver behind these results.
Laura Perrott, Global Director of Brand & Digital Marketing, Colt
Get started to connect your stack and turn what your customers are already telling you into your next campaign.
Frequently asked questions
Is there an official Intercom MCP server?
Yes. Intercom runs an official remote MCP server at mcp.intercom.com, maintained by Fin (the company formerly called Intercom, Inc.). It supports OAuth 2.0 or Bearer-token authentication and is available for US and EU hosted workspaces, with AU not yet supported.
Is Intercom now called Fin?
The company behind Intercom renamed itself Fin on 12 May 2026, and agreed to a roughly $3.6 billion acquisition by Salesforce on 15 June 2026, expected to close in Q4 of Salesforce’s fiscal 2027. The help desk product marketers and support teams know as Intercom keeps its name; only the parent company and the AI agent are called Fin.
What can marketing teams do with the Intercom MCP server?
Pull verbatim customer language, feature requests and objections straight out of support conversations into an AI conversation, for campaign briefs, positioning work or content, instead of exporting tickets or asking support to summarise a theme manually.
Is the Intercom MCP server secure?
It inherits your existing Intercom permissions and supports OAuth with admin oversight, but any MCP server that reads support conversations surfaces whatever the connected account can see, including sensitive customer data, so scope tokens carefully and prefer read-only access where you can.
Is the Intercom MCP server free to use?
Connecting the official server itself is free. It runs on top of whatever Intercom or Fin plan and seats your workspace already pays for, so there’s no separate MCP fee, though write actions still count against your normal API usage.