CLI Overview
The Semantius CLI is a single binary, semantius, that gives you full access to your Semantius platform from a regular shell. Where the MCP connectors target interactive AI chats and the agent skills target IDE-based coding agents, the CLI is the layer underneath: a scriptable, no-frills entry point built for automation and headless agents.
What the CLI Is For
Use the CLI whenever you need Semantius to run without a human in the loop:
- Headless agents running on a server, in a container, or in the background, that need to introspect the model, read or write business data, and run analytical queries without a chat client wrapping them.
- Shell scripts and Bun / Node scripts that chain Semantius calls together (read an entity, transform with
jq, write a record, kick off an analytical query). - Recurring or scheduled jobs that seed data, run nightly imports, or fan out an analytical query across many tables.
- One-off tasks in the terminal: search tools, inspect a tool’s schema, or call a tool with inline JSON.
The CLI works in two modes: crud for managing your model and reading or writing records, and cube for CubeJS-compatible analytics. Any tool you can call from Claude or ChatGPT through the MCP connectors, you can call from a shell, with the same arguments.
Why a CLI in Addition to MCP
The MCP connectors are perfect for conversational use, but they assume an MCP host on the other end. The CLI removes that assumption:
- No client required. Run it from bash, zsh, PowerShell, a Docker entrypoint, a cron job.
- Composable. Pipe output through
jq,xargs,awk, save to a file, fan out, retry, branch. - Predictable. Stable command surface (
info,grep,call), JSON in and JSON out, exit codes you can check. - Fast. A built-in lazy-spawn daemon keeps MCP connections warm between calls so chained commands do not pay startup cost every time.
Installation
The CLI is installed as part of the Agent Skills setup, since the skills call the local semantius binary under the hood. Follow the installer for your OS, then verify with semantius --version:
The same page covers credential setup (SEMANTIUS_API_KEY, SEMANTIUS_ORG) via either shell exports or a .env file. Once that is done, the CLI is ready to use on its own, with or without an agent.
Next Steps
- The semantius command →: learn the basic command surface, the
crudandcubemodes, and how to call them. - use-semantius skill →: drop-in skill that teaches an agent to use the CLI productively.