Dashboard

MCP Integration

Definition. Noxtica exposes a read-only Model Context Protocol (MCP) server. It lets your own external AI agents read Noxtica — policies, rules, alerts, risk distribution — over JSON-RPC, authenticated with scoped, rate-limited, audited bearer tokens.

Two facts to fix up front, because they define what this is and isn’t:

  • Read-only. Your agents can read what Noxtica knows. They cannot write to Noxtica, change configuration, or act through it.
  • Opt-in per tenant. The MCP server is available, but it is not on by default. A tenant enables it deliberately, behind a default-off flag. There is no self-serve “connect now” — it is something you turn on when you want it.

How Noxtica does it

MCP is the open standard for giving an AI agent structured access to a tool or data source. Noxtica’s server speaks JSON-RPC and is the integration meaning of “agentic”: your agents reading us.

Authentication

Access is mediated by bearer tokens that an operator mints, with three properties:

  • Scoped — a token grants access only to the surfaces you intend, not blanket access.
  • Rate-limited — calls are throttled, so an agent can’t hammer the server.
  • Audited — every call is logged, so you have a record of what your agents read and when.

The read-only tool set

The server exposes tools that read Noxtica state, for example:

  • Policies — read your configured policies.
  • Rules — read the rules driving decisions.
  • Alerts — read recent alerts.
  • Risk distribution — read how risk is distributed across your traffic.

Every tool is a read. There is no write tool. An external agent connected over MCP can build context about your Noxtica posture — to summarize, to monitor, to reason about — but it cannot make a change, and it cannot act on your behalf through Noxtica.

Where it fits

MCP is the “you integrate agentically” meaning on the agentic spine. It is sharply distinct from Know Your Agent (which governs which agents reach your site) and the AI Assistant (which helps you operate the console). See Agentic Security for how the three fit together.

Read deeper