Is Claude secure for your meeting notes? What happens when you connect Granola

TL;DR: Connecting Granola to Claude via MCP gives Claude access to your enhanced notes and, on Business or Enterprise plans, full transcripts. Device audio is never shared: Granola deletes it immediately after transcription. Basic plan access is limited to the last 30 days of notes with no transcripts. Business and Enterprise unlock full history. What Claude does with that data depends on your Claude plan tier: consumer plans (Free, Pro, Max) follow consumer retention and training policies, while commercial plans (Team or Enterprise) carry no training on inputs or outputs by default under commercial terms. Enterprise workspaces require an admin to enable MCP before users can connect.

Back-to-back meetings create a documentation gap, and capturing what was said is only the first half of the problem. Finding it months later, when someone asks what was decided, what was promised, or what you learned, is where most note-taking systems fail. The connection between Granola and Claude through the Model Context Protocol (MCP) turns your meeting notes into a queryable archive. The connector shares your enhanced notes and, on Business or Enterprise plans, full transcript access, but it does not give Claude your audio files.

The natural question when connecting two tools is what each one can see. The answer depends on which Granola plan you use, which Claude tier you connect with, and how sensitive the conversations in your archive are. This article explains exactly what moves between the systems, what stays private, and how to configure the connection to match the kind of work you do.

What moves between Granola and Claude, and what doesn't

When you connect Granola to Claude through MCP, you establish a channel where Granola acts as the MCP server and Claude acts as the MCP client. Claude can request your notes and, on paid plans, a meeting's full transcript, within the scopes you approve and the active workspace.

This is deliberate: Granola built and maintains the MCP connector directly, so authorised requests, access scopes, and revocation controls all work as described.

How MCP data flows to Claude

When Claude needs context to answer a question like "Why are enterprise customers hesitant about SSO?" it sends a request to Granola's MCP endpoint. Granola returns only the specific notes or transcript snippets relevant to that query across all your past interviews.

Device audio never travels through this pipe. Granola transcribes device audio in real time, then deletes the source audio immediately. This protects participant privacy by ensuring device audio never persists beyond the transcription process.

Only your enhanced notes and, depending on your plan, raw transcripts remain available for MCP queries. You authenticate individually, so Claude can only access notes you are authorized to see.

Granola's first-party connector gives you revocation control that manual copy-paste workflows can't match. For a detailed walkthrough of the connection setup, see the Granola MCP tutorial which demonstrates the connection flow step by step.

How Claude handles your meeting data

Claude processes data differently depending on whether you use the consumer chat interface or connect through the API and MCP. This is different from consumer chat interactions where you paste text directly into Claude's interface.

Consumer Claude accounts (Free, Pro, or Max) follow different retention and training policies than commercial accounts (Team or Enterprise). If you connect Granola to a consumer Claude account, those consumer policies apply. Connecting to a Claude Team or Enterprise account gives you commercial terms, which by default do not use inputs or outputs to train models.

How participant data stays private

Granola's privacy model protects participant data through architecture, not policy alone:

• Granola transcribes device audio in real time and immediately deletes the source file, as described above.
• Your notes are yours, taken from your point of view, capturing what matters to your work.
• Claude queries only what you authorize through connection settings, which you manage from Claude's integrations page.
• You can delete specific segments of a transcript if a participant shares sensitive feedback, preventing those segments from being accessible via MCP.

Data privacy: What you get on each plan

Granola and Claude both use tiered approaches to data access. Understanding where these align helps you configure the connection safely.

What Basic plan users can query over MCP

Granola's Basic plan limits MCP access to your last 30 days of meeting notes. You cannot retrieve full transcripts on Basic. Queries return enhanced notes only. For many research workflows, 30 days is enough to synthesize recent discovery interviews. If you need to query patterns across months or quarters of customer research, Business or Enterprise plans remove the time limit.

Enterprise workspace controls

On Granola's Enterprise plan, MCP is disabled by default at the workspace level. Individual users can see the setting but only admins can change it. To enable MCP, an admin navigates to Settings → Workspace → General → Apps & connectors → MCP access → Manage. If you're evaluating Granola for customer research and your company uses Enterprise, you'll need to ask your admin to enable MCP before you can connect Claude.

Admins can configure specific access scopes, restricting whether Claude can read notes only or also access transcript data. For organizations with strict compliance requirements, this creates a clear approval chain before any customer interview data flows to external AI tools.

You control what Claude can access and can revoke it anytime

The connection between Granola and Claude uses a standard authorization protocol that lets you control access without sharing your Granola password.

What Claude can access

When you initiate the connection, you authorize specific permissions through a browser-based flow. The connector requests access to your notes and, on paid plans, transcripts. You can revoke access at any time from Claude's connection management page.

Disconnecting Claude from your research archive

In Claude's settings, you can review all active MCP connectors under the connections or integrations section. Each connector shows which service it connects to.

To disconnect Granola from Claude, navigate to Claude's connections settings and select Disconnect. This removes Claude's authorization to query your Granola data.

How Granola and Claude handle data security

Security certifications and data handling policies determine whether connecting Claude complies with company requirements and protects participant trust.

SOC 2 Type 2 compliance

Granola is SOC 2 Type 2 certified as of July 2025. The full SOC 2 report and security documentation are available through Granola's Trust Center. Anthropic's commercial products also maintain SOC 2 Type 2 audited status and GDPR compliance.

SOC 2 Type 2 audits the operational effectiveness of security controls over time, not just at a single point. Granola's certification covers its own systems and infrastructure. Claude Enterprise's SOC 2 Type 2 certification covers Anthropic's systems.

Data hosting and regional storage

Granola stores all data on Amazon Web Services (AWS) servers in the United States. Granola does not currently offer EU data residency. Granola encrypts your data in transit and at rest.

For Claude, Enterprise customers have access to custom data retention, customer-managed encryption keys, and US-only inference. If your organization has specific data handling requirements, verify current availability with both vendors before connecting.

Enterprise access controls

Granola's Enterprise plan supports SSO/SAML and SCIM provisioning, giving IT teams centralised control over who can access the workspace and what they can do within it. SSO/SAML lets employees log in through your existing identity provider (such as Okta, Azure AD, or Google Workspace) so access is automatically revoked when someone leaves the organisation. SCIM provisioning automates user lifecycle management, syncing seat additions and removals directly from your directory without manual intervention.

Together, these controls mean access to your meeting notes archive, and by extension any Claude MCP connection, follows the same governance policies as the rest of your company's software stack.

How to handle sensitive feedback in your research archive

Customer interviews often contain confidential product feedback, competitive intelligence, or personal participant experiences. Here is how to handle that data safely when using Granola with Claude.

Removing sensitive interview segments

If a participant shares sensitive feedback such as competitor names, unreleased product details, or personal health information, you can use Granola's delete-parts-of-transcript feature to remove specific segments while keeping the rest intact. Deleted segments are no longer accessible via MCP.

Interview storage and deletion

Granola doesn't keep audio after transcription. You can delete individual notes or entire meetings from Granola at any time. Deleted notes become inaccessible via MCP.

Managing third-party data permissions

Granola's third-party AI providers (the services Granola uses on the backend to power transcription and note enhancement) are contractually prohibited from training on your data. This is a separate relationship from your own Claude account connecting to Granola via MCP.

When you connect your personal or team Claude account through MCP, that connection is governed by your own Anthropic plan terms, not by Granola's subprocessor agreements. Under commercial API terms, API inputs are not used for training by default.

For Enterprise users, commercial terms explicitly prevent training on customer data. Consumer users should verify current terms directly with Anthropic.

How to tell participants Granola is transcribing

Granola's transparency features make it straightforward to inform participants that transcription is occurring. Before connecting Claude to your customer research workflow, check your organization's consent policy.

Transcription consent requirements

Consent requirements for transcribing conversations vary by location. Before conducting interviews, check the consent laws that apply to you and your participants.

Granola's transparency features are designed to make disclosure straightforward. The automatic chat notification sends a message at the start of the call so all participants know transcription is occurring. The video watermark displays a visible indicator throughout the session, including for anyone who joins late. Both features work during MCP-connected sessions and help you build a consistent disclosure practice.

Starting the interview by letting participants know you are using a transcription tool is good practice. Once you turn them on, Granola's chat notification and video watermark handle this automatically, so disclosure does not depend on you remembering to say something.

Before you connect: A checklist

Claude Team and Enterprise for research workflows

Claude Enterprise operates under commercial terms, which guarantee no training on customer inputs. Enterprise admins also receive audit logs and additional controls.

If you use Claude's consumer tier (Free, Pro, or Max) with Granola, consumer terms apply, which may include different retention and training policies unless you opt out. For customer research containing confidential feedback or participant data, check Anthropic's usage policies and your organization's guidelines to confirm which Claude tier meets your requirements. The distinction matters: consumer and commercial accounts follow different terms, so verify directly with Anthropic what applies to your situation.

Research interviews vs. internal meetings

Internal team meetings and customer research interviews require different privacy postures. Internal sprint planning or 1-on-1s might be fine on Basic plans with 30-day limits and consumer Claude tiers. Customer discovery interviews containing competitive intelligence, unreleased product feedback, or participant personal data need stricter controls.

Before connecting Claude to Granola for customer research, use this checklist:

1. Verify your plan tier: Ensure you are on Business or Enterprise if you need to query your full history, and confirm your admin has enabled MCP if you're on Enterprise.
2. Check your Claude tier: Connect Granola only to Claude Team or Enterprise accounts for customer research to guarantee no-training terms.
3. Confirm consent practices: Verify you're informing participants about transcription using Granola's transparency features.
4. Test the query functionality: Run a test query across a few past interviews to confirm the research repository value before committing to the workflow.

Try Granola for free. Download the Mac or Windows app, connect your calendar, and run your next meeting to see it in action.

FAQs

Is Claude secure for customer research interviews?

Granola's MCP connector shares only enhanced notes and, on Business or Enterprise plans, full transcripts. Device audio is never shared. For Claude's own data handling, retention, and training policies, refer to Claude's Trust Center. For customer interviews containing confidential feedback, use Claude Team or Enterprise and inform participants about transcription using Granola's transparency features.

Can I use Claude with Granola if my company has security requirements?

Granola is SOC 2 Type 2 certified and GDPR compliant, and Claude Team and Enterprise maintain SOC 2 compliance. Granola currently stores data only in the US. If your company has specific security requirements, security teams can review Granola's Trust Center and evaluate whether the platform meets your requirements. Enterprise workspace users should confirm MCP is enabled before connecting.

What happens to my customer research if I disconnect Claude?

Disconnecting removes Claude's authorization to query your Granola data. Your research notes remain in your Granola workspace, and the archive persists even when you stop using Claude. To restore access later, you will need to re-authorize the connection. For questions about note retention after disconnection, refer to Granola's MCP documentation.

Does connecting Claude violate participant consent?

Granola's transparency features give you a straightforward way to inform participants that transcription is occurring. Use the chat notification or video watermark at the start of the interview, and follow your organization's consent protocols. For guidance on your specific consent obligations, consult your legal team or your organization's compliance guidelines.

How do I query across multiple customer interviews?

Once you've connected Granola to Claude via MCP, you can ask questions that span your entire research archive. For example, "Why are enterprise customers hesitant about SSO?" or "What UX issues came up most often in Q1 discovery calls?" Claude queries your Granola notes and returns answers. Basic plan queries cover recent notes only. Business and Enterprise plans let you query your full history.

Key terms glossary

MCP (Model Context Protocol): A protocol that lets AI tools like Claude request specific data from applications like Granola using authorized, scoped permissions.

OAuth: An authorization protocol that lets you grant Claude limited access to your Granola notes without sharing your password. You approve specific permissions (scopes) and can revoke them anytime.

Connection permissions: The specific access granted to Claude, defining which categories of data the AI can request from your Granola workspace.

Data residency: The geographic location where your data is stored. Granola currently stores data on AWS servers in the United States only.