@voltade/volty
v0.1.6
Published
Remote CLI for the Volty platform — run operations, trigger real agent wakes, and inspect LLM I/O over an API key.
Readme
@voltade/volty
The Volty command-line tool for Claude Code. You set it up once; after that, you don't use it directly — Claude does. It lets Claude test the Volty agent for real (safely, in a sandbox — nothing ever reaches a real customer) and look inside what the agent saw and decided.
One-time setup
Open a terminal and run these four steps.
1. Install Bun (skip if you already have it):
curl -fsSL https://bun.sh/install | bashThen close and reopen your terminal.
2. Install the Volty tool:
bun add -g @voltade/volty3. Teach Claude Code how to use it:
volty skill install4. Log in — this opens your browser; approve it with your normal Volty account:
volty auth login --env prodTo check it worked: volty auth whoami --env prod should print your email, org, and role.
That's it. You never need to memorize any other commands.
Using it — just ask Claude
Open Claude Code anywhere and ask in plain language. The skill you installed in step 3 kicks in automatically. For example:
- "Test the Volty agent — ask it whether we're GST registered."
- "Send the agent a follow-up in that same test conversation."
- "What did the agent actually reply with, and why? Show me what the LLM decided."
- "Why didn't the agent learn anything from that conversation?"
- "Clean up the test conversations on prod."
Claude knows the safety rules: tests run against a sandboxed draft version of the agent with a synthetic contact, so nothing can ever be sent to a real customer, and Claude cleans up after itself when it's done.
Upgrading
When a newer version exists, the tool prints a notice — Claude will see it and upgrade for you. To do it yourself:
bun add -g @voltade/volty@latest && volty skill install