MCP connector

Connect KollGuard to Claude and other AI clients

KollGuard runs a remote MCP server. Add its URL to an MCP client, sign in to KollGuard, and approve read-only access. The assistant can then read your findings, posture and scans and answer from live data.

Server details

Server URLhttps://api.kollguard.com/v1/mcp
TransportStreamable HTTP
AuthorizationOAuth 2.1 with PKCE (S256) and dynamic client registration. No client ID or secret to enter.
Scopemcp:read (read-only)
TokensAccess token 1 hour, refresh token 30 days
AccessPer user, limited to that user’s KollGuard organization

You need a KollGuard account. The server shows only what your account can already see in KollGuard.

Add it to Claude

Claude on the web and Claude Desktop

  1. Open Settings, then Connectors, and choose Add custom connector.
  2. Paste https://api.kollguard.com/v1/mcp as the remote MCP server URL. Leave the OAuth client ID and secret empty.
  3. Click Add. You are sent to KollGuard to sign in, then to a screen that lists the access being granted.
  4. Click Allow. KollGuard appears in your connectors list.
  5. In a chat, ask for example “Use KollGuard to show my current security posture” or “List my open critical findings from KollGuard.” Claude asks before each tool call.

Claude Code

claude mcp add --transport http kollguard https://api.kollguard.com/v1/mcp

Then run /mcp inside Claude Code and choose KollGuard to sign in and approve access.

Add it to other clients

Any MCP client that supports remote servers over streamable HTTP with OAuth can use the same URL. Where a client takes a config file, the entry is usually a URL and nothing else; the file name and top-level key differ by client, so check the client’s own MCP documentation.

{
  "mcpServers": {
    "kollguard": {
      "url": "https://api.kollguard.com/v1/mcp"
    }
  }
}

The client opens a KollGuard sign-in the first time it connects. For a client that cannot do OAuth, or for a local setup, use a read-only API key with the kollguard-mcp package instead; see the developer docs.

Tools

The server offers eight tools. Each is read-only and is marked as such in its tool annotations (read-only hint set, destructive hint off).

ToolWhat it does
kollguard_list_projects
List projects
Your KollGuard scan targets (id, name, type). Call this first in an organization with several projects.
kollguard_list_findings
List findings
Security findings. Optional filters: projectId, severity (low, medium, high, critical), status, framework (for example SOC2, HIPAA).
kollguard_get_posture
Get security posture
Open findings, weighted risk score, per-framework gaps and top risks.
kollguard_list_scans
List scans
Recent scan runs (GitHub and database) with per-severity counts and status.
kollguard_stats
Get headline stats
Open findings by severity and framework, and scans in the last 7 days.
kollguard_list_issues
List issues
Issues in the Developer Tools tracker. Optional filters: epicId, status.
kollguard_list_epics
List epics
Epics, the parent containers for issues.
kollguard_list_tickets
List support tickets
Support tickets. Optional filters: status, category.

What the connector cannot do

  • It cannot create, change or delete anything. There are no write tools.
  • It cannot start scans, change settings, or read another organization’s data.
  • The AI client you connect receives what the tools return and handles it under that client’s own terms.

See Security and the Privacy Policy for how KollGuard handles data.

Frequently asked questions

Can the KollGuard connector change anything in my account?

No. All eight tools are read-only. Each one reads data and returns it. The connector is granted a single read scope (mcp:read), so it cannot write, delete or start a scan.

Whose data can the connector see?

Only the organization of the person who approved it. Access is per user and scoped to that organization. It sees the same findings, posture and scan results as the KollGuard dashboard.

How do I remove the connector?

Remove KollGuard from the connectors (or MCP servers) list in your AI client. That ends its access. Access tokens last one hour and refresh tokens 30 days, so deleting the connector in the client is the way to stop it at once.

Do I need an API key?

Not for the remote connector: you sign in and approve access. API keys (kgr_) are for the local npx server and the REST API, described in the developer docs.

Need help? Email hello@kollguard.com. More developer docs are at /docs