vela mcp config
Print the server entry an AI client needs to start vela mcp.
Synopsis
Print the server entry an AI client needs to start vela mcp. Nothing is written: each client keeps its own configuration format, so the entry is printed either as the mcpServers JSON block most clients accept, or as the server's own command and arguments, which go after a client's "mcp add".
The entry points at the absolute path of this vela binary and carries the server flags given here. The server uses the kubeconfig and context of the client process that starts it, so start the client from a shell pointed at the cluster you mean.
vela mcp config [flags]
Examples
# JSON block to paste into a client's MCP configuration
vela mcp config
# Claude Code, read and plan tools, restricted to one namespace
eval "claude mcp add --scope user kubevela -- $(vela mcp config --format command --namespace team-a)"
# Allow writes, with confirmed direct mode only in dev
vela mcp config --allow-write --direct-namespace dev
Options
--allow-write Register apply_application. Without --direct-namespace it only returns YAML for a pull request. Off by default.
--direct-namespace stringArray Allow direct mode, confirmed by the user in their client, in this namespace. Repeat for more than one. Refused at start if it matches --production-pattern.
--format string Output format: json for an mcpServers block, or command for the server command and arguments a client's mcp add takes. (default "json")
-h, --help help for config
--name string Server name in the JSON block. (default "kubevela")
--namespace stringArray Restrict Application, resource and log reads to this namespace, including resources an Application manages elsewhere. Repeat for more than one; empty means any namespace RBAC allows. Definitions, addons and clusters in the system namespace stay readable.
--production-pattern string Regular expression for production namespace names, which never allow direct mode. (default "(?i)(^|[-_.])prod(uction)?($|[-_.])")
Options inherited from parent commands
-V, --verbosity Level number for the log level verbosity
-y, --yes Assume yes for all user prompts
SEE ALSO
- vela mcp - Serve KubeVela to AI agents over the Model Context Protocol.