Vulnix
API

MCP (AI apps)

Connect Claude, ChatGPT or any app that supports the Model Context Protocol to your Vulnix workspace, and drive pentests and findings in plain language.

Vulnix runs a hosted MCP server. Add it to an AI app, sign in with your Vulnix account, choose what the app may do, and then ask it things like "start a pentest on the staging scope" or "which critical findings are still open?".

https://api.vulnix.dev/mcp

Connecting an app

  1. In your AI app, add a custom connector (a remote MCP server) with the URL above.
  2. The app opens a Vulnix sign-in page. Sign in, pick the workspace, and tick what the app may do.
  3. Select Allow. You're sent back to the app, connected.

Nothing is installed and no token is copied. The app gets its own connection, which you can see and disconnect under API → MCP in the console.

Permissions

The permissions are the same scopes as API tokens. A connection can never do more than the person who connected it: if their role changes, the connection changes with it. Permissions that use up pentests (starting pentests, re-testing fixes, opening fix PRs) are never ticked for you.

If the app later needs a permission it wasn't given, it asks you to reconnect with that permission added. Your AI app will also ask you to confirm before it runs any action that changes something.

What an app can do

Every tool maps to an existing API route, with the same rules, limits and Free-plan restrictions.

AreaTools
Pentestslist, get, progress, agent trace, findings, compare two runs, start, stop, wait for a run to finish
Findingslist (by status and severity), get, history, update status or accept risk, check a fix, open a fix PR
Overviewsecurity score, open findings over time, remediation time
Domainslist, get, findings and pentests per domain, test accounts, add, verify, remove
Scopeslist, get, create, targets, surface classification, constraints, context, repository, schedule
Knowledgelist, add, edit, delete
Repositories and PR reviewslist and read, PR review settings, PR review findings
Organizationupdate profile, list members, invite
Billingplan and usage, plans, pentest history
Searchsearch and fetch across findings, pentests and domains

Chat and webhooks are console features and are not available to AI apps.

Workspace rules

Owners and admins control AI apps under API → MCP:

  • Allow AI apps to connect. Turning this off stops every connection immediately.
  • Limit what apps can do. A ceiling on the permissions anyone can grant. Existing connections are narrowed at their next request.

Admins also see every connection in the workspace and can disconnect any of them.

Agents without a browser

A CI job or a headless agent can't complete a sign-in page. It can call the same server with an API token as a bearer header instead:

Authorization: Bearer vnx_...

The token's scopes decide which tools work.

Limits

Each tool call counts against the connection's rate limits, the same as an API call: 120 per minute, and 10 per minute for actions that start a pentest or re-test a fix.

On this page