Skip to main content
Before you start, create a key and store it as described in Local AI agents. The examples assume QUIVA_API_KEY is set in your shell.

Install the plugin

The QuivaWorks plugin bundles the remote MCP connection and the quiva-config skill in one install. In Claude Code:
The first command registers the repository as a marketplace. The second installs a plugin as <plugin>@<marketplace>. From a shell the same steps are claude plugin marketplace add QuivaWorks/mcp and claude plugin install quiva@quivaworks. After it is active, type / and look for the plugin’s skills under the quiva name.

Or add the remote server yourself

The shell expands $QUIVA_API_KEY when you run the command, so the key is stored in your Claude Code configuration. Add --scope user to use it in every project. The default, local, is only you in the current project, and project writes a shared .mcp.json. Run /mcp in Claude Code to confirm quiva is Connected.

Share it in a project

To share the setup with your team without sharing a key, commit a .mcp.json that reads the key from each person’s environment:
Claude Code replaces ${QUIVA_API_KEY} from your environment when it loads the file.

Local stdio servers

To run the local servers with a profile instead of one .env file per server, pass the profile name:
Repeat for each server you want. --env takes KEY=value pairs and needs another option, such as --transport stdio, between it and the server name. The server finds its key as described in Local AI agents.

First prompts

Start with something that changes nothing: “List my spaces”, “Which flows are published?” Anything that writes or deletes is refused once and needs confirm: true before it runs.