Palette MCP Server
Model Context Protocol (MCP) servers allow AI models to interact with external tools, data sources, and services in a structured and reliable way. Models that are connected to MCP servers can perform actions such as read files, query databases, or call APIs; they do not have to rely on custom logic authored by you or code dynamically generated by the (Large Language Model) LLM itself. AI tools connected to MCP servers are useful in practical applications because they become active assistants, with capabilities far beyond generating text.
The Palette MCP Server wraps around the Palette API, allowing you to use natural language to perform actions on resources. The MCP server is a powerful automation tool that enables LLMs to interact with Palette consistently without having to understand the complete Palette API.
The Palette MCP server ships as a container image and as a native binary for macOS and Linux, and lets you interact with the following Palette resources:
- List, filter, and read status and observability data for clusters, retrieve their kubeconfig files, along with the option to create, update, and delete them.
- Read and update cluster tags.
- List, filter, create, and delete cluster profiles.
- List, create, update, and delete projects, teams, and users.
- List and filter packs, registries, cloud accounts, and edge hosts.
- Read component events per cluster, per edge host, or tenant-wide.
The server operates in read-only mode by default. Write tools, such as create, update, and delete, are gated behind the
--allow-write startup flag. Delete tools additionally require a typed-name confirmation. For the full tool catalog,
refer to Palette MCP Server Operations.
Check out the tutorials in the AI section to learn how to incorporate the Palette MCP server into your workflows.