Kiali MCP

Expose Kiali capabilities to AI assistants using the Model Context Protocol (MCP).

Kiali MCP is an integration that allows MCP-capable AI assistants to query (and optionally manage) Kiali-related data by calling tools exposed by an MCP server.

The implementation is provided as part of the Kubernetes MCP Server upstream and also for Openshift MCP server. It exposes a kiali toolset (see upstream guide: docs/KIALI.md).

Prerequisites

  • A reachable Kiali endpoint (Route/Ingress/Service URL).
  • Kubernetes credentials available to the MCP server (kubeconfig or in-cluster config).

Enable the kiali toolset

Create a TOML config file and enable kiali in toolsets.

toolsets = ["core", "kiali"]

[toolset_configs.kiali]
url = "https://kiali.example" # Endpoint/route to reach the Kiali console
# insecure = true  # optional: allow insecure TLS (not recommended in production)
# certificate_authority = "/path/to/ca.crt"  # CA bundle for Kiali's TLS cert

Notes:

  • If url is https:// and insecure = false, you must provide certificate_authority.
  • Authentication to Kiali is performed using the server’s Kubernetes credentials (it obtains/uses a bearer token for Kiali calls).

Connect from an MCP client

How you wire this into a specific client depends on the client, but the core idea is the same: start the MCP server with your kubeconfig and your TOML config.

Example (conceptual) command:

kubernetes-mcp-server --config /path/to/config.toml --read-only

Once connected, your assistant can use the Kiali tools (for example: mesh graph, metrics, traces, workload logs) to power a chatbot-like experience outside the Kiali UI (for example, in an IDE).

Last modified March 3, 2026: [AI] Add AI section (#955) (962121c)