MCP Security& Usage
How AI assistants connect to X-Analytics through the Model Context Protocol, and what that access allows.
The X-Analytics MCP (Model Context Protocol) server lets approved AI assistants — such as Claude — securely work with your organization's cyber-risk data on your behalf. Access is granted only with your explicit consent, is scoped to your own organization, and can be revoked by you at any time.
This page covers the X-Analytics MCP server that AI clients connect to. For the inbound MCP connectors that bring CrowdStrike, Tenable and Drata data into X-Analytics, see MCP Connectors.
Connecting
Each organization connects on its own dedicated URL, available inside the product under Account Settings → Connectors:
https://{your-org}.app.x-analytics.com/mcp
When you add the connector in your AI client, you'll be taken to an X-Analytics consent screen showing exactly what access is being requested. The connection is established only after you approve it.
Setting Up the Connector
This walkthrough takes you from a standing start to your first tool call. You connect with your own X-Analytics account — there is nothing to install and no credentials to request in advance.
Before you start
- An X-Analytics account, and an account owner who can approve the connection.
- Your organization's MCP URL, which you find in step 1 below.
- Two-factor authentication is enforced on every X-Analytics account. If you have not enrolled yet, the app walks you through it on first sign-in and shows a QR code you can scan with any authenticator app.
- An AI client that supports remote MCP servers over Streamable HTTP — for example Claude, or the MCP Inspector for testing.
Step by step
Find your organization's MCP URL
Sign in to X-Analytics and go to
Account Settings → Connectors. Each organization has its own subdomain, so only ever use the URL shown inside your own account. It takes this form:https://{your-org}.app.x-analytics.com/mcp
Add the connector in your AI client
In Claude, open
Settings → Connectors → Add custom connector, paste the URL and click Connect. In the MCP Inspector, select the Streamable HTTP transport and paste the same URL. No credentials need to be issued beforehand — your client registers itself through OAuth 2.1 dynamic client registration.Sign in and approve access
You are redirected to X-Analytics to sign in, then to a consent screen listing exactly what is being requested: which risk profiles, and which of the three permissions — Read cyber risk, Read cyber maturity reports, and Set framework scores. Grant only what you need; permissions are enforced on the tool list itself, so tools behind a permission you did not grant are never shown to the assistant and cannot be called. The connection is established only after you approve it.
Confirm the connection
Ask your assistant to list your risk profiles. It should call
xa_list_profilesand return the profiles on your account. Note the one you want to work with — every other call is scoped to a profile.Try it
These prompts exercise different parts of the toolset:
- “What is our cyber risk for [profile name]?”
- “How do our threat scores compare to our industry?”
- “What are our top risk-reduction opportunities, and what would each one save us?”
- “Show our NIST CSF 2.0 maturity and where the biggest gaps are.”
- “Compare our current exposure with the inherent baseline.”
Updating framework scores
Two tools can write: xa_set_nist_csf_2_scores and xa_set_cis_8_1_scores. Both require the Set framework scores permission, both are annotated destructive, and your AI client will ask you to confirm before either one runs.
Submissions are partial by design — only the entries you include are updated. Asking your assistant to set NIST sub-category GV.OC-01 to 3, for example, changes that one sub-category and nothing else. You can read the result back with xa_get_nist_csf_2_scores.
Troubleshooting
Limits and controls
- Rate limit: 600 requests per minute per integration.
- Revoke at any time in
Account Settings → Connectors. Revocation takes effect immediately, and no support request is needed. - Connector activity is recorded in an audit log tied to the authorizing user.
Authentication & Access
- Connections to the X-Analytics MCP server authenticate with OAuth 2.1 — there is no API-key or password path, and your X-Analytics credentials are never shared with the AI client.
- A connected assistant receives the same access as the user who authorized it, limited to that user's organization. It cannot see or act on any other organization's data.
- Access tokens are short-lived and bound to the approved connection.
What the Connector Can Do
The server exposes 14 tools: 12 that only read data and 2 that write. Every tool carries an accurate read/write annotation, so your AI client can tell you what it does before it runs.
Which tools appear depends on the permissions you grant during consent. A permission you do not grant means its tools are never listed and cannot be called — without write:framework-scores, for instance, 10 of the 14 tools are visible.
Profiles
Listed on every connection. This tool needs no permission beyond being signed in.
| Tool | Access | What it does |
|---|---|---|
xa_list_profiles | Read | The risk profiles available on your account. |
Read cyber risk
read:cyber-riskCyber risk totals, baseline, trends, threat scores, loss scenarios and risk-reduction opportunities for the profiles you share.
| Tool | Access | What it does |
|---|---|---|
xa_get_cyber_risk | Read | Current financial cyber exposure for a profile, broken down by category, asset group and threat. |
xa_get_cyber_risk_baseline | Read | Inherent, pre-control exposure next to the mitigated figure. Computed on demand, so slower than the other reads. |
xa_get_cyber_risk_trends | Read | How exposure, control effectiveness and threat scores have moved over time, monthly and quarterly. |
xa_get_threat_scores | Read | Threat scores for a profile, including the comparison against your industry. |
xa_get_loss_scenarios | Read | Modelled loss scenarios behind the exposure figure, with probability and impact for each. |
xa_get_risk_reduction_opportunities | Read | Ranked actions that reduce exposure most, with the projected reduction from each. |
Read cyber maturity reports
read:cyber-maturityNIST CSF 2.0 and CIS Controls v8.1 maturity reports and overall control effectiveness for the profiles you share.
| Tool | Access | What it does |
|---|---|---|
xa_get_nist_csf_2_report | Read | NIST CSF 2.0 maturity report: overall alignment and tier, by function and category. |
xa_get_cis_8_1_report | Read | CIS Controls v8.1 maturity report: overall alignment and implementation group, by function and control. |
xa_get_control_effectiveness | Read | Overall control effectiveness for a profile, with the per-control breakdown across the frameworks you use. |
Set framework scores
write:framework-scoresSubmit, update and read NIST CSF 2.0 and CIS Controls v8.1 framework scores. This one permission gates all four tools, including the two that only read.
| Tool | Access | What it does |
|---|---|---|
xa_get_nist_csf_2_scores | Read | Reads the NIST CSF 2.0 sub-category scores already submitted for a profile. |
xa_get_cis_8_1_scores | Read | Reads the CIS Controls v8.1 safeguard maturity already submitted for a profile. |
xa_set_nist_csf_2_scores | Write | Submits or updates NIST CSF 2.0 sub-category scores. Partial submissions leave every other entry untouched. |
xa_set_cis_8_1_scores | Write | Submits or updates CIS Controls v8.1 safeguard maturity. Partial submissions leave every other entry untouched. |
Nothing else in your account can be changed through the connector. The two write tools only set framework maturity scores, only on profiles you already have permission to edit, and only for the entries you include.
Tool inventory last verified . The definitive list is always the one your AI client shows after connecting.
Managing & Revoking Access
You stay in control of every connection, with no support request required. To review or revoke a connected assistant, go to Account Settings → Connectors and disconnect it. Revocation takes effect immediately — active sessions stop working right away.
Data & Privacy
- Data is encrypted in transit and at rest, and is kept isolated per organization.
- Information retrieved through the connector is sent to the AI assistant you choose and is then processed by that assistant's provider under its own terms and privacy policy, which X-Analytics does not control. Please review what you ask the assistant to retrieve.
- X-Analytics does not use data accessed through the MCP server to train AI models.
For how we handle personal data more generally, see our Privacy Policy.
Only connect to the official X-Analytics server on app.x-analytics.com, and only use AI clients you trust. If you are unsure whether a connector is genuine, confirm the URL inside your account before approving access.