You can connect Fathom to Claude using the Model Context Protocol (MCP), and once it’s connected, Claude can search your meeting recordings, pull up a transcript, or summarize what was decided on a call without you digging through Fathom’s own search first. Fathom’s OAuth also just works with a new client, unlike several other tools in this series. This guide covers both ways to connect it — directly or through a gateway — and what’s worth checking before an AI client can read every meeting your team has recorded.

What the Fathom MCP server actually does

Fathom publishes a remote MCP server at api.fathom.ai/mcp. It exposes your recorded meetings — transcripts, summaries, and action items — as tools Claude can call directly, so prompts like these work against your real meeting history:

  • “What did we agree on the call with [client] last week?”
  • “Find every meeting where pricing came up this month.”
  • “Summarize the action items from yesterday’s standup.”
  • “Who was on the call where we discussed the renewal?”

That’s genuinely useful — a meeting transcript is exactly the kind of thing people mean to search later and rarely do, because scrolling back through recordings is slow. It also means the server’s tools can surface a full transcript, word for word, including whatever a client, candidate, or teammate said on a call they may not have expected an AI tool to read later.

Option 1: Connect Fathom to Claude directly

Claude supports remote MCP servers natively. The short version of the setup:

  • Open Claude’s connector or MCP settings.
  • Add a remote server pointing at Fathom’s MCP endpoint.
  • Sign in with your Fathom account when prompted — standard OAuth, so Claude never sees your Fathom password, and you can revoke access from Fathom’s own account settings at any time.

That’s enough if Claude is the only AI client you use with Fathom. The tradeoff shows up once a second client or a second teammate needs the same access: each one repeats the same OAuth grant, and each grant is its own thing to remember later.

Option 2: Connect it once, through a gateway

If you’re already juggling more than one MCP server or more than one AI client, wiring each pair directly gets old fast — the connection sprawl covered in what an MCP gateway is. The idea is to connect Fathom (and everything else) to one gateway URL, then point every AI client at that single URL instead of repeating the OAuth step per client.

With gate specifically, connecting Fathom looks like this: add gate’s gateway URL to Claude once, then pick Fathom from the Fathom server page in the directory and sign in with OAuth the same way you would directly. From then on, Fathom’s tools are available in every client connected to your gate URL — Claude, ChatGPT, or Cursor — with the same per-tool allow, ask, or block rule applied everywhere instead of set up separately in each client. See the full list of servers set up this way on the MCP servers page.

What to check before you connect a meeting recorder

Most tools in this series expose your own data. A meeting recorder exposes other people’s words too — clients, candidates, and teammates who were on the call but never agreed to have it searched by an AI tool afterward. A few things worth checking:

  • Which meetings does the grant cover? Confirm whether the connection can see every recording on your Fathom team, or only calls you personally hosted or attended, before you ask a broad question like “summarize everything from this quarter.”
  • Read, or read and manage? Check the actual tool list rather than assuming it’s search-only — gate’s free MCP security scanner lists every tool a server exposes, so you can see whether anything on there can delete or share a recording rather than just retrieve one.
  • Who else on the team can reach it? A shared, revocable connection through one gateway beats every teammate holding their own grant to the same recordings — the access-control question covered in MCP access control for teams.
  • Could something said on a call steer what Claude does next? A transcript is content someone else spoke, not data you authored — the same untrusted-input pattern covered in the confused deputy problem in MCP. An agent that already has a legitimate connection to your meeting history doesn’t need a malicious server to misuse it, just one line in a transcript it wasn’t expecting to act on.

Troubleshooting the connection

  • OAuth loop or sign-in fails. Make sure you’re signing in with the Fathom account that’s actually on the team whose recordings you want Claude to see, not a personal login for a different workspace.
  • Claude can’t find a meeting you know exists. Fathom’s tools respect your permissions in Fathom itself — if a recording belongs to a team or folder you don’t have access to, Claude won’t see it either.
  • Tools disappear after a Fathom update. Remote servers can change their tool list without warning. If you’re connected through a gateway, that’s exactly what drift detection is meant to catch — see how MCP rug pulls work for why a one-time review isn’t enough.
Where gate fits: gate connects Fathom once, behind a URL every AI client can share, with allow/ask/block rules you can set per tool — so a transcript search can run on its own while anything that shares or deletes a recording waits for a person. Every server is scanned for prompt injection before it goes live, and every call goes into a plain-language log, which matters more for a shared meeting archive than for most connections in this series. See the full approach at /mcp-gateway.

The bottom line

Connecting Fathom to Claude with MCP is a normal OAuth flow, direct if it’s your only client or once through a gateway if you want the same access shared across every AI client your team uses. The useful part — searching months of meetings in one prompt — is also the part worth a slower look, since the transcripts it searches were spoken by people who weren’t necessarily thinking about an AI tool reading them back later.