Tools

What You'll Build

Three connected pieces that turn every conversation your team has into structured, searchable, actionable data:

  1. Fathom records every meeting. Zoom, Google Meet, Microsoft Teams, phone calls, even in-person. Transcript, summary, action items, the whole thing.
  2. Claude reads the entire archive. Through Fathom's MCP integration, you can ask Claude things like "what did Sarah agree to in our March demo?" or "what objections came up the last five times we pitched the enterprise tier?" It answers from the meetings themselves, not from your memory.
  3. Fathom Workflows handles the after-the-meeting work. Every recording triggers automatic actions: pain points written to the CRM, follow-up email drafts in your inbox, deal-blocker flags posted to Slack. You finish a call and the busywork is already done.

The combination is the part nobody else covers. Most companies have Fathom (or Otter, or Gong). They use it as a note-taker. That's like buying a Bloomberg terminal and only using the news ticker.

Why This Works

The most expensive part of every sales call, customer success conversation, and prospect interview is not the call. It's everything that has to happen afterward:

The math on this is brutal. A sales rep doing five calls a day spends 30โ€“45 minutes per call on the after-work. That's three hours every day. Fifteen hours every week. Three quarters of a work day per week, per rep, lost to post-meeting admin.

Fathom alone solves the "writing the summary" part. That's roughly 10 minutes of the 30. The other 20 minutes โ€” the CRM update, the email, the Slack message, the recall six months later โ€” has historically required either a human or a sprawling Zapier mess.

The fix is the three-piece stack. Fathom captures. Claude recalls. Workflows acts. Each one is a single SaaS subscription. None of them require code.

How the Three Layers Work

Layer 1: Fathom Captures Everything

Fathom is a meeting recorder. You install it once and it auto-joins every call you have, in the background, without needing to invite a bot to the meeting. For every call it produces:

Every meeting goes into your library forever. Searchable by keyword, attendee, date, deal, or custom tag.

Layer 2: Claude Reads the Whole Library

Fathom's MCP server (announced October 2025) is what makes the library queryable from Claude. Once it's connected, Claude has access to:

This is the part that's underused. Most Fathom customers ask their notes "what happened yesterday." With the MCP connected, you can ask what happened across every meeting you've ever recorded:

Claude answers from the actual transcripts, not your memory.

Layer 3: Fathom Workflows Does the Work

A Workflow in Fathom is "when this happens in a meeting, do that automatically." It's the after-the-call automation layer. Every workflow has a trigger and one or more actions:

Triggers:

Actions (built-in):

A real workflow looks like this. Trigger: a new sales call. Actions: write the pain points to the HubSpot deal's "Pain Points" field, write the budget signal to "Budget Stage," post the action items to #sales-deals in Slack, draft the follow-up email in Gmail and stage it for review. Total human work after the call: zero, until you open Gmail to review and send the email.

Step-by-Step Setup

Step 1: Connect Fathom to Your Calendar

Install Fathom. Connect your work calendar (Google Workspace or Microsoft 365). Fathom now sees every meeting on your calendar and joins each one in the background. You can set rules for which meetings get recorded (internal only, external only, specific contacts, specific calendars).

There's no bot in the meeting โ€” Fathom records via the meeting platform's native API, not by sitting in the call as an invisible attendee. Cleaner experience for everyone on the call.

Step 2: Pick Your Summary Templates

Fathom lets you choose or build the summary template per meeting type. Start with the built-in ones:

Fathom auto-routes meetings to the right template based on calendar metadata.

Step 3: Wire the Fathom MCP into Claude

Fathom has a hosted MCP server at https://api.fathom.ai/mcp. To connect Claude Code:

  1. Run claude mcp add fathom https://api.fathom.ai/mcp (or the equivalent for Claude Desktop).
  2. The first time you query it, Claude opens a browser tab for OAuth. Sign in with your Fathom account.
  3. Verify the connection by asking Claude "how many meetings did I have last week?"

Now Claude can list, search, and read any meeting in your library. The whole archive is one conversation away.

Step 4: Build Your First Workflow

In Fathom, go to Workflows. The first one to build is the sales-call automation. Configure:

Test it on one real sales call before turning it on for all of them. Verify every field landed in the right place.

Step 5: Add Workflows for Each Meeting Type

Once the sales-call workflow is solid, build the next ones:

The pattern is the same every time. New recording โ†’ extract structured fields โ†’ push to the right destinations.

Step 6: Use Claude for the Long-Tail Questions

The workflows handle the predictable post-meeting work. Claude handles the unpredictable questions:

These are the questions that don't fit a templated workflow. Claude reads across the library and answers in seconds.

Step 7: Train the Team on the Question Habits

The hardest part of this whole setup is not technical. It's behavioral. Most people will keep using Fathom only to summarize the call they just had. Getting your team to ask "what does the meeting library say about this" before every customer touchpoint is the actual leverage.

Two things help:

  1. A shared prompts collection. Fathom lets you save and share canned prompts. Save the ones that work for your team: "Top 5 objections this quarter," "All-time mentions of ," "Customers who churned and what they said in their last call." Make them one-click for everyone.
  2. A weekly meeting-library digest. A small workflow that, every Friday, asks Claude to summarize the week's meetings across the team and posts to a Slack channel. Reading the digest gets people in the habit of querying the library.

Adapting This for Other Conversation-Driven Businesses

The same three-layer stack works for any business where conversations are the unit of value.

Consulting firms. Every discovery call gets summarized into a structured intake doc and pushed to the project management tool. Claude becomes the institutional memory across decades of client engagements.

Recruiters and search firms. Every candidate call gets pushed into the ATS with structured fields. Claude searches across years of candidate conversations to surface re-engagement opportunities.

Customer support teams. Every support call gets a workflow that updates the ticket, tags the product feature, and posts a deep link to the bug if one was reported. Claude finds repeating issues across calls.

Medical and legal practices. Every consult gets summarized into the relevant case file (with HIPAA-compliant tooling). Claude searches across the practice's history for similar prior cases.

Real estate agents. Every showing call gets logged into the CRM with the buyer's stated must-haves and dealbreakers. Claude finds the right inventory match against years of buyer conversations.

Wherever your business runs on conversations, this stack turns those conversations into data you can search and act on.

Gotchas and Tips

What This Replaces

Before this stack:

After this stack:

One vendor. One conversation. Every meeting your business has ever had, awake and useful.


Keep Reading