# The AI meeting assistant that actually works with Claude

Most AI meeting assistants solve the wrong problem.

They transcribe. They summarize. They send you a tidy document when the call ends. And then the decisions in that document sit in a folder you never open again.

Two weeks later, you are in Claude explaining the same context you covered in that meeting. Claude does not know what you decided. It does not know what you agreed to. It starts from zero.

The transcription was not the problem. The missing layer is what happens after.

## Why most AI meeting assistants fall short

Tools like Otter, Fireflies, Fathom, and Granola are built around the call.

They help you capture what happened.

But Claude does not need another transcript. It needs reusable context.

That means decisions, tradeoffs, owners, and the reason behind the decision. Saved somewhere Claude can read later.

## What Claude actually needs from your meetings

Claude is not in your meetings. It has no idea what you discussed, what you committed to, or what changed.

Claude needs three things from your meeting:

- The decision that was made
- The context behind it
- The next steps

Not a transcript. Not a summary no one will read. Three pieces of information in a place Claude can actually access.

## The meeting-prep template

Hjarni is a Markdown knowledge base with a built-in MCP server. It ships with a [meeting-prep template](/templates/meeting-prep), a structured note with five sections:

**Before the meeting**
- What is the goal of this meeting?
- What context does everyone need?
- What decision needs to be made?

**During the meeting**
- Key points discussed
- Decisions made

**After the meeting**
- Next steps and owners
- What changed from what we expected?
- What should Claude know about this?

The last section is the one that matters. It is the bridge between your meeting and your AI.

## How it works with Claude and ChatGPT

Connect Hjarni's built-in MCP server. It takes five minutes. ([Claude guide](/docs/connect-claude-mcp), [ChatGPT guide](/docs/connect-chatgpt-mcp)) For the integration overview, see [Hjarni for Claude](/for/claude).

After a meeting, fill in the template. Save it to your Meetings folder.

The next time you open Claude and work on anything related to that project, Claude reads the meeting note. It knows what was decided. It knows the context. It does not ask you to repeat yourself.

You ran a product review with your team. You decided to cut a feature and ship sooner. You noted that the client is sensitive about delays.

A week later you ask Claude to draft a client update. Claude reads the meeting note. The draft reflects the actual decision. You do not explain the context again.

That is the difference.

## One meeting note beats ten follow-up chats

Most people use AI to compensate for poor documentation. They paste in context at the start of every session. They re-explain the background. They summarize previous decisions inline.

A single well-structured meeting note, saved to Hjarni, replaces all of that.

Claude reads it when you ask. You stop rebuilding the context by hand.

Write once. You both remember.

## Not a transcription tool. Not a summary tool.

Hjarni is not trying to join your calls. It does not record anything. It is not another tab open during a meeting.

It is what you reach for after. Five minutes to capture what matters. Then let Claude carry it forward.

## Set it up

1. Create a free Hjarni account at [hjarni.com](https://hjarni.com)
2. [Connect Claude](/docs/claude) or [ChatGPT](/docs/chatgpt) to Hjarni's built-in MCP server
3. Use the [meeting-prep template](/templates/meeting-prep) for your next meeting
4. Ask Claude: "Read my latest meeting note for this project before you answer."

The context stays. You stop re-explaining.

Want to skip the manual step? You can pipe transcripts from Otter, Fireflies, or any tool into a Hjarni note with [Zapier](/docs/zapier), then keep them short and structured from there. See [5 ways to automate your knowledge base](/blog/automate-knowledge-base-with-chatgpt-and-zapier) for more of those workflows.

Meetings are usually one connector among several. For how this memory layer sits beside your calendar, CRM, and the rest, see [combining multiple MCP servers](/blog/combine-multiple-mcp-servers).

[Free to start. No credit card required.](/docs/getting-started)