Quick start for AI users

Vulnara has an MCP server, so an AI assistant such as Claude or ChatGPT can read your workspace and answer questions about it: which repositories have critical findings, what the last scans found, how your security score is moving. This guide connects your assistant and gets you a first answer.

You need a Vulnara account that belongs to at least one workspace, and an AI client plan that allows custom connectors.

1. Copy the connection details

   You need two values:

   • MCP URL: https://vulnara.rso.dev/mcp
   • Client ID: hl04e6MSMRY60LdpGh5rdMRQjkPxvldAYoqXdzo4

   Neither is a password. The client ID is the same for every customer. Access is protected by your own Vulnara sign-in, and you never paste a token.

2. Add Vulnara to your AI client

   Claude (claude.ai and Claude Desktop): open Settings, then Connectors, and choose Add custom connector. Paste the MCP URL. Open the advanced settings, choose to use your own OAuth client, and paste the client ID. On Claude Team and Enterprise, an Owner adds the connector once and each member then signs in individually.

   ChatGPT: turn on developer mode, then create an app with the same MCP URL and client ID.

   Claude Code and other MCP clients: see MCP.

3. Sign in with your Vulnara account

   Your client opens the Vulnara sign-in page. Sign in with the same account you use for the web app, and approve the request. The assistant now acts as you, with your role in each workspace, and never with more.

4. Choose a workspace

   If you belong to one workspace, there is nothing to choose. If you belong to several, ask the assistant which workspaces you have, then tell it which one to use. It lists them with list_workspaces and passes the one you pick to every other tool.

5. Ask a question

   Try one of these:

   • "Which of my repositories has the lowest security score?"
   • "Summarise the dashboard for the last 30 days."
   • "List the critical and high findings in the repository called payments-api."
   • "What did the most recent scan of my main repository find?"
   • "Show me the triage decisions taken on this repository, newest first."
   • "Which of my findings are in the CISA KEV catalogue?"
   • "Which risks in the register are overdue for review?"

   The assistant should say which workspace it read from. If the answer looks wrong, check that first.

What the assistant can and cannot do

• Mostly read: almost every tool reads. A few can start a scan, triage a finding, cancel a task, change a repository's settings or work a risk in the risk register, and only if your AI access is Write. Nothing deletes data, and nothing admin-only, financial or about people is offered. See MCP.
• Your permissions only: the assistant sees what you can see in the web app, checked on every call. If your role changes, the next call uses the new role.
• Never exposed: git tokens, service accounts, billing and invoices, and the raw values of secrets that scanners found. Credential-shaped values are removed from results.
• Untrusted text is marked: file paths, repository names and similar text come from your repositories, and anyone who can commit to them controls it. The server marks that text as data, so a well-behaved assistant does not follow instructions hidden in it.
• Large results are paged: long lists come back a page at a time. Ask the assistant to read further pages if it stops early.

The full list of tools is in the MCP reference.

Chat in the web app

You do not need an AI client to ask a question. When you are signed in to the web app, a chat button sits in the corner of every screen. Open it and type a question.

• No separate sign-in: the chat uses your web app session, so it has your account and your role, and nothing more.
• One workspace at a time: it answers about the workspace you have selected, and keeps a separate conversation for each workspace. Switching workspace switches the conversation. Start a new one with +.
• The same tools as MCP: it looks up your data with the same tools as the MCP server, so the same data stays out of reach.
• Changes run straight away: if your AI access is Write, the chat can start a scan, triage a finding and similar when you ask it to. It does so directly, without a confirmation step, so phrase requests for changes deliberately. With Read, it can only look things up.
• A daily allowance: each workspace has a daily chat allowance. Once it is used up, the chat answers with links to the documentation pages that best match your question instead of a written answer, until the allowance renews.
• Replies stream in: the answer appears as it is written. Replies have a Copy button, and a reply that arrives while the chat is closed is marked on the chat button.
• Your size, your window: drag the chat's top-left corner to resize it; the size is kept in your browser. Pop it out into its own window and put it back, and the conversation comes with it. Escape closes it, and on a phone it fills the screen.
• Several lines: Shift+Enter starts a new line in a message, Enter sends it.
• Signed-in members only: the chat is not shown before you sign in.

Next steps