AI assistants are changing how cybersecurity, IT, and GRC teams find information and complete everyday work. But without access to the systems where that work actually happens, these AI assistants are limited to the context a user provides.

The now fully hosted Secureframe MCP Server enables you to provide tools like Claude Code, Claude Desktop, and Cursor with permission-aware access to your compliance data so you get better answers and work more effectively, without needing to run any infrastructure yourself.

That means in a few clicks you can start asking your AI assistant questions in natural language, getting answers grounded in current Secureframe data, and allowing it to take actions on your behalf based on your existing permissions, all without switching tools. 

What’s possible with the Secureframe MCP server?

With the Secureframe MCP Server, you can ask your AI assistant to surface failing controls, investigate high-risk vendors, or check user compliance status across frameworks like CMMC, NIST 800-171, FedRAMP, and SOC 2. You can also prompt it to take action on your behalf to update a test status, add a comment to request evidence, or create a POA&M item, for example.

Common tasks using the MCP server include:

• Checking compliance and framework progress
• Finding and updating issues in security controls or compliance tests
• Reviewing personnel and device status
• Investigating third-party and vendor risk data
• Managing risks and POA&M items, including creating and updating entries
• Reviewing and updating SSP sections, assessment objectives, policies, and roles
• Responding to Trust Center requests

“This is a powerful step forward for our customers. We’re helping them tap into their compliance data with the AI tools they already use, without requiring them to host, monitor, or update any infrastructure themselves,” said Shrav Mehta, Founder and CEO of Secureframe. “It’s all about making compliance easier, more accessible, and more integrated with your daily workflow.”

How does the Secureframe MCP Server work?

The Secureframe MCP Server implements the Model Context Protocol (MCP), an open standard for connecting AI assistants to external systems. This allows clients like Claude Code, Claude Desktop, and Cursor to query and act on your compliance data based on simple natural language prompts like “show me failing SOC 2 controls.”

The Secureframe MCP Server exposes more than 100 tools across 38 categories. Each tool wraps a Secureframe REST API endpoint, so the query filters and parameters available on an endpoint are available on the matching tool. Connected clients receive each tool’s parameter schema automatically. This allows an assistant to construct the right request from a plain-language question.

Some examples are listed below.

Browse the complete catalog, organized by category and labeled read or write, in the Secureframe MCP Server docs.

How do permissions work?

You can connect a client if you’re an Admin or Super Admin, your role allows you to create API keys, or you already have an active API key. 

Every request runs as the user tied to the connection, and is limited to that user’s existing Secureframe permissions. If a user cannot see or change something in Secureframe, an assistant working on their behalf cannot either.

That behavior is worth considering before you connect a client. The server supports the same write actions as the API, including creates, updates, and deletes. An assistant working on your behalf can therefore change real data. If you’re connecting with an API key, generate one scoped to a user whose permissions match the tasks you want the assistant to perform.

What’s the benefit of the Secureframe MCP Server being fully hosted?

Secureframe hosts the MCP server, so there is nothing for customers to install, deploy, monitor, or update. This makes it easier to connect an AI assistant without taking on another service to maintain.

The hosted server provides region-specific endpoints for US and UK workspaces. For most clients, you point your MCP client at the endpoint and sign in with your existing Secureframe login, with no key or secret to generate, copy, or store. The first time the client connects, it opens your browser, you sign in and approve access, and the connection is ready to use. 

You get the same 100+ tools across 38 categories, the same read and write coverage as the Secureframe REST API, and the same permission model tied to your Secureframe account.

Five ways teams are using the Secureframe MCP Server

By making compliance data instantly accessible through natural language queries, and actionable through the same interface, the Secureframe MCP Server makes everyday workflows faster, simpler, and more aligned with how cybersecurity leaders and developers actually work.

Below are some of the most powerful ways Secureframe customers are using the MCP server:

1. Spot failing controls before they become audit findings

Asking for failing controls filtered by framework or owner. The assistant can pull the surrounding context, including which tests feed the control, who owns it, and what changed. You get a sense of why something is failing, not just a list.

2. Surface high-risk vendors for faster reviews

Querying vendor data directly for high-risk vendors that require immediate review. Turn vendor reviews into a working queue instead of a spreadsheet exercise, making ongoing monitoring more realistic to sustain.

3. Streamline remediation with test insights and task management in one place

Pulling a real-time view of failing tests, then moving straight into action to accelerate preparation for an assessment and stay ready. You can ask “What are the most recent failing tests in our CMMC environment?” and follow up with “Assign myself as the owner of this test.” Because the lookup and the action happen in the same place, findings are less likely to sit and go stale in a document waiting for someone to act on them.

4. Identify inactive users and mitigate access risks

Querying personnel records to find inactive users, contractors with lingering access, or accounts that may need to be offboarded. AI assistants help flag these risks and support access reviews by surfacing the most relevant personnel data automatically.

5. Automate actions in Secureframe without switching tools

Asking the assistant to pull information or take action in Secureframe without needing to switch between tools. For engineers especially, this is often the most useful part. If you’re in Cursor or Claude Code and need to add a comment to a control in Secureframe requesting evidence before shipping a change, you can do so without leaving your current workflow.

How to connect your AI assistant to the Secureframe MCP Server

Getting started takes four steps:

Step 1: Generate an API Key and Secret 

Admins and Super Admins can sign in directly without this step. All other users need an API key or the ability to create one before they can connect, even with OAuth. To generate a key, go to Company Settings → API Keys. 

Step 2: Choose the endpoint for your region.

US: https://mcp.secureframe.com/

UK: https://mcp-uk.secureframe.com/

Step 3: Add Secureframe as an HTTP MCP server 

Add the Secureframe MCP Server in Claude Code, Claude Desktop, Cursor, or another MCP-compatible client. Most clients support standard OAuth and will prompt you to sign in and approve access the first time they connect. If your client can’t open a browser, supply your API Key and Secret in the Authorization header instead, separated by a space:

Authorization: <API_KEY> <API_SECRET>

Step 4: Restart the client

After restarting the AI assistant, you can begin querying your Secureframe data. MCP servers load when a client starts, so the connection will not appear until you restart.

For the exact commands for Claude Code, Claude Desktop, Cursor, and other MCP clients, consult the Help Center setup guide. 

How to use the MCP server in your workflows securely

As you start exploring and acting on your compliance data, follow these steps:

• Always verify that insights from AI assistants align with your organization’s security standards and compliance policies.
• Review what the assistant is about to do before confirming, especially for updates, deletes, or archive actions.
• If you’re connecting with an API key, scope it to a user with the permissions needed for the tasks you want the assistant to perform.
• Store your key and secret in environment variables rather than pasting them into a shared config file.

A smarter, more connected future of cybersecurity

Secureframe is built on the belief that compliance should work with your tools, not against them. The Secureframe MCP Server brings us closer to that goal by creating a secure, standardized way to connect AI assistants to the systems that power your cybersecurity program.

Whether you’re preparing for an assessment, reviewing test results with your team, or monitoring and correcting drift, the ability to access and understand your data in context makes all your cybersecurity efforts faster and more intuitive.

We’re continuing to expand how Secureframe connects to the tools your team already uses, including plugin support for MCP clients. In the meantime, you can connect your client and try a question you’d normally answer by clicking through the platform. 

For more information, read the documentation and follow the setup guide, or talk to an expert to learn more about how Secureframe can simplify your cybersecurity program end-to-end.