Meeting Automation

How to Use Granola AI and MCP to Give AI Agents Context from Your Meetings

Your meeting history as context for every AI assistant and agent, through one MCP connection.

GranolaMCPAI Agents

By Troy Tessalone · · 9 minutes

Meeting Automation

A practical field guide from Automation Ace.

Short answer: Granola offers a hosted MCP server at https://mcp.granola.ai/mcp that lets AI assistants and agents such as Claude, ChatGPT, Claude Code, and Cursor search your meetings, read notes, and (on eligible plans) pull transcripts. Add Granola as a connector in Claude or ChatGPT, or add the server URL in other MCP clients, then sign in through your browser; there is no API key. Your AI can then prep you for calls, draft follow-ups, and answer questions across months of meetings, with the meeting dates to back them up.

Tool names, setup paths, and plan limits come from Granola and third-party guides at the time of writing and vary by source.

Short answer

  1. Connect Granola in Claude or ChatGPT, or add mcp.granola.ai/mcp to your MCP client.
  2. Sign in with OAuth; no keys or config secrets.
  3. Ask questions across meetings, prep for calls, draft follow-ups.
  4. Check plan limits for history and transcripts.
  5. Keep sensitive meetings out of scope.

What the Granola MCP server does

MCP (Model Context Protocol) lets AI applications use tools from other services. Granola's MCP server exposes your meetings as tools: listing and searching meetings, reading notes, querying in natural language, and retrieving transcripts. The AI decides when to call them while answering you. Background: how AI agents use tools and MCP.

How to connect it

ClientHow to connect (at the time of writing)
ClaudeSettings → Connectors → add Granola, then sign in
ChatGPTSettings → Apps → connect Granola, then sign in
Claude Code, Cursor, other MCP clientsAdd the remote server URL https://mcp.granola.ai/mcp, then authorize in the browser
Your own agentUse an MCP client library with the same URL and OAuth flow
// Example MCP client config for tools that use a JSON file
{
  "mcpServers": {
    "granola": { "url": "https://mcp.granola.ai/mcp" }
  }
}
// The client opens a browser sign-in; no API key in the file.

What an assistant can ask

CapabilityExample question
List and search meetings“Which meetings did I have with Acme this quarter?”
Read meeting notes“Summarize what we agreed in yesterday's pricing call.”
Natural-language queries across meetings“What objections have prospects raised about onboarding?”
Transcripts (plan-dependent)“Quote exactly what the client said about the deadline.”

Use cases

Use casePrompt or agent task
Meeting prep“Brief me for my 2 p.m. with Acme: last 3 meetings, open commitments, risks.”
Follow-up drafting“Draft a follow-up email for today's call with the agreed next steps.”
Account history“What has Acme asked for across all our calls this year?”
Product feedback“List feature requests from customer calls last month, grouped by theme.”
Hiring“Compare my notes from the three candidate interviews against the scorecard.”
Engineering contextIn Cursor or Claude Code: “Use the requirements from Tuesday's spec review.”

Building agents with meeting context

Agent instructions (meeting-prep agent, example)

Before each external meeting on the calendar:
1. Use Granola to find previous meetings with the same company or attendees.
2. Summarize: relationship history, commitments we made, open questions, risks.
3. Check the CRM for deal stage and recent activity.
4. Post a 10-line brief in Slack to the meeting owner 1 hour before.

Rules: cite the meeting date for every commitment. Read-only. If no prior
meetings exist, say so in one line.

Trigger it from your calendar or a booking; see triggering an AI agent from an API or webhook and automating bookings. Combining sources: pair Granola with your CRM and web research, for example research agents with Firecrawl.

Limits, privacy, and accuracy

ConsiderationDetail
Plan limitsSources describe the free plan as limited to recent notes and no transcripts; paid plans extend access. Check current plans
ScopeThe assistant sees the notes your Granola account can access; enterprise admins can control permissions
Data flowContent you retrieve goes to the AI provider you are using (for example Anthropic, OpenAI, or Cursor)
Sensitive meetingsKeep HR, legal, and confidential notes out of reach or out of automated prompts
AccuracyAsk the assistant to cite meeting dates; check quotes against transcripts

Agent safeguards: permissions, approvals, and audit trails.

MCP vs integrations vs Zapier

  • MCP: on-demand questions and agent context; nothing happens until someone or something asks.
  • Native integrations: push notes to Slack, Notion, or your CRM after meetings.
  • Zapier: event-driven workflows (tasks, field updates, email drafts) when notes are filed; with Zapier, start from Granola's triggers.

Build the event-driven side: from meeting notes to action with Granola and 5 Granola follow-up automations.

Frequently asked questions

Does Granola have an MCP server?

Yes. Granola offers a hosted MCP server at https://mcp.granola.ai/mcp that lets MCP-capable assistants such as Claude, ChatGPT, Claude Code, and Cursor search meetings, read notes, and on eligible plans retrieve transcripts.

How do I connect Granola to Claude or ChatGPT?

In Claude, add Granola from Settings, Connectors. In ChatGPT, connect it from Settings, Apps. Then sign in through your browser. Other MCP clients use the server URL with the same sign-in.

Do I need an API key for the Granola MCP server?

No. The Granola MCP server uses browser-based OAuth sign-in, so there is no API key to paste into a config file.

Can the Granola MCP server access transcripts on the free plan?

Sources at the time of writing describe the free plan as limited to recent notes without transcripts, with paid plans extending access. Check Granola’s current plans.

Is it safe to connect meeting notes to an AI assistant?

Connect only accounts and meetings that should be reachable, keep sensitive meetings out of scope, remember that retrieved content goes to the AI provider you use, and ask the assistant to cite meeting dates for accuracy.

GranolaMCPAI Agents

Disclaimer: Granola features, integrations, plan limits, and MCP tools change, and recording rules vary by location; details reflect public documentation and coverage at the time of writing, so confirm current details with each provider before building. Automation Ace is not affiliated with Granola beyond the referral links marked on this page. This article may include links to apps, products, or services; some links may be affiliate links, which means Automation Ace may earn a commission at no extra cost to you.

Build Better Systems

Ready to automate with confidence?

Share your tools, process, and goals. Automation Ace can design the Zapier, AI, and automation setup that fits your business.

Start a Project →