Astra MCP security, permissions and data

Last updated: November 2, 2026

This article explains how Astra MCP handles requests, permissions, sign-in and data.

How a request flows

  1. The user types a request in the AI tool, for example "List the open critical vulnerabilities on my production target".

  2. The assistant chooses one or more Astra MCP tools, such as target_list and then vulnerability_list, and sends the request to https://mcp.getastra.com/mcp.

  3. Astra MCP checks the user's sign-in and calls the Astra platform as that user.

  4. Astra returns the data. Astra MCP passes it back to the assistant.

  5. The assistant writes the answer in the chat. For actions such as starting a scan, the assistant first asks for any missing choice, then confirms what changed.

Most AI tools show which tool is being called and ask the user to approve it before it runs. The exact prompt depends on the AI tool.

Permissions

  • The assistant acts as the signed-in Astra user. It can only see and change what that user can see and change in the Astra dashboard.

  • No API key is needed. The connection uses the same Astra sign-in as the dashboard.

  • Some actions depend on the plan and on the user's role in Astra. For example, the report types and scan types offered through Astra MCP are the same ones offered in the dashboard for that account.

Read, write and delete actions

Each tool is labeled Read, Write or Delete. Some AI tools also use these labels to decide when to ask for approval.

Label

What it means

Examples

Read

Looks up data. Nothing changes in Astra.

vulnerability_list, scan_get, report_get, scan_validate_start

Write

Creates or changes something in Astra.

scan_start, comment_add, rescan_request, report_generate, vulnerability_apply_transition

Delete

Removes or stops something.

scan_cancel, comment_delete, schedule_delete

For the full list of tools and their labels, see Astra MCP tools reference.

Sign-in

Astra MCP uses browser sign-in (OAuth). No password or API key is stored in the AI tool; only the server URL is saved there. The first sign-in asks the user to review and allow access. For the steps, see Connect Astra MCP to your AI tool.

Tip

Only approve the Allow access page when the sign-in was started from the user's own AI tool. After the first sign-in, the AI tool renews access in the background, so the sign-in is not repeated every session. If the saved sign-in stops working, the AI tool asks to sign in again.

What data is read and written

  • Read: targets, scans and scan statistics, vulnerabilities, findings and their evidence, comments, rescans, schedules, reports, activity log entries, workspaces, companies, subscriptions and CI/CD integrations that the user can access in Astra.

  • Written: new scans, scan cancellations, vulnerability workflow actions, comments, rescan requests, schedules and report requests. Each change is made with the user's own account, the same as a change made in the dashboard.

  • Not handled by Astra MCP: report file downloads, CI/CD integration setup or tokens, and HAR file uploads.

Important

AI output can be wrong. Review the assistant's answer before acting on it, especially before starting a scan, changing a vulnerability status or deleting a schedule.

For anything not covered here, contact help@getastra.com.