VICIPanel
Docs/MCP server/MCP security and audit
MCP server

MCP security and audit

How authentication, authorization, rate limiting, and audit logging work for MCP.

5 min read

Connecting AI to your dialer is powerful, it's also a new attack surface. VICIPanel's MCP implementation takes security seriously. Here's what you get out of the box.

Authentication

  • OAuth 2.1 with Dynamic Client Registration (PKCE-protected), recommended for team use
  • Bearer tokens with mcp_ prefix, simpler, good for personal or CI use
  • Both methods respect the user's VICIPanel permission level at every tool call

Authorization

Every tool call is authorized against the caller's user level. An Agent-level token cannot run Manager-level tools, cannot list other agents' data, cannot modify campaigns. If a tool needs Admin (L9), only an Admin token can run it.

Rate limiting

Standard tools: 60 requests per minute per token. Destructive tools: 10 per minute. Heavy compute tools (report generation): 5 per minute. Exceeding limits returns a standard 429 response with Retry-After header.

Session and token lifecycle

  • OAuth session TTL: 30 minutes idle, 8 hours maximum
  • Concurrent sessions per user: 100
  • Bearer token expiry: configurable per token at generation time
  • Revocation: instantaneous, revoked tokens stop working within seconds

Audit log

Every MCP request writes a structured audit entry: user, token, timestamp, tool called, arguments (PII redacted), result status, and response time. Admins can search, filter, and export the audit log for compliance reviews.

Don't share personal tokens

A personal token is tied to your user. If you share it with a teammate, their actions look like yours in the audit log. Use OAuth with DCR for team access, each user carries their own identity.

Want to see this in action?

The live demo runs on real data. Instant credentials, no signup required.