TaruviBase Agent Plugin
Official TaruviBase plugin for Claude Code, Cursor, Codex, GitHub Copilot, Gemini CLI, Kiro, and other AI coding agents. It gives your agent:
- the TaruviBase MCP server (
https://mcp.taruvibase.com/mcp) to manage your app's tables, records, functions, policies, storage, secrets, and queries taruvi-app-developerfor backend worktaruvi-refine-providersfor Refine frontend worktaruvi-setupto connect a project and write its.env
Getting started
Install the plugin for your agent. The first time your agent uses a TaruviBase tool, it opens your browser so you can sign in and approve the sites and apps it may use. There is no API key to paste into MCP config.
| Agent | Install |
|---|---|
| Any supported agent | npx plugins add Taruvi-ai/taruvibase-plugin |
| Claude Code | /plugin marketplace add Taruvi-ai/taruvibase-plugin, then /plugin install taruvi@taruvi |
| Codex | codex plugin marketplace add Taruvi-ai/taruvibase-plugin, then install from the Plugins Directory |
| GitHub Copilot CLI | copilot plugin install Taruvi-ai/taruvibase-plugin |
| Gemini CLI | gemini extensions install https://github.com/Taruvi-ai/taruvibase-plugin |
| Kiro | Powers → Add Custom Power → Import power from GitHub → repository URL |
Then ask your agent:
setup taruvi
It confirms the MCP connection and writes your app's .env from the values on the
app's Settings → Connect page in TaruviBase Console.
Starting a new web app? You can clone the TaruviBase Refine starter template (Refine v5 + React + MUI, providers pre-wired), or use any framework you like.
If you already added the TaruviBase MCP server by hand, remove it after installing the plugin so your agent doesn't load the same tools twice.
Repository structure
See AGENTS.md. In short: skills/ is shared by every agent, and
each agent has a thin manifest that points at it and at the shared MCP server.
Develop
npm test