New Agent CLI & API

Hydra's CLI: Manage GPU Fleet from Terminal

When AI engineers live in the terminal shell - their GPU fleet management should, too. The Hydra CLI puts our full GPU control plane in the terminal where AI Platforms already work. Every operation that used to require a dashboard is now one command, and for teams using AI coding agents, the CLI gives AI agents a direct way to interact with your GPU fleet.

Your Terminal is home to everything - except your GPU infrastructure

SSH, Ansible, shell scripts, Python, Kubernetes, Claude Code - they all run from the command line. The terminal is where engineers build, debug, and automate. Every layer of the stack is one command away.
 
But GPU infrastructure management has required context switching and reaching for browser-only interfaces that don't talk with your other tooling. Need to check server availability? Open a browser. Provision new capacity? Click through a multi-step UI. Power cycle a node? Dashboard again. Each operation pulls you out of the terminal environment.
 
For teams running AI coding agents - tools like Claude Code that can write code, run tests, and deploy services - there's no easy way to plug them into infrastructure operations.

Agent-Ready GPU Infrastructure

We built the Hydra CLI so agents like Claude Code can talk to your GPU infrastructure. Install via NPM, run brokkr login, and you're ready to go.
 
We built a specific command for AI agents: docs. When an agent runs it, the CLI outputs the complete reference - every command, every flag, every JSON output schema. The agent reads this once and knows how to interface with your entire Hydra GPU fleet.
 
Here is what that looks like in practice. You ask your AI coding agent to check GPU health across a deployment. The agent runs brokkr deployments --json to list your active servers, selects the target, extracts the IP and connection string, SSHes into the server, runs nvidia-smi, and reports back GPU health across your fleet. All from a simple prompt. No dashboard, no clicking through menus.
 
This is about making it easy for engineers to bring AI into GPU infrastructure workflows - quick answers about deployments, status, and health from a simple prompt.

Automation Ready

The Hydra CLI slots directly into the automations engineers already run - Ansible playbooks, shell scripts, Python. Check server status while tailing logs. Power cycle a node between Ansible tasks. Reprovision a server from a shell script. Every deployment operation is a single command with structured output, ready to plug into your existing toolchains.
 
Deployment operations give you direct control over your servers. List all deployments with sorting, filtering, and search. View full details on any deployment by ID. Power on, off, or cycle via IPMI. Activate rescue mode. Lock deployments to prevent accidental destructive actions. Reprovision, decommission, and organize servers into projects - each a single command.
 
Organization management covers the operational layer your platform team runs day-to-day. Control members, roles, and invitations. Create and manage API keys. Set up webhooks with full delivery history and retry. Handle billing and payment methods. Manage org settings - all without opening a browser.
 
Pipe deployment lists into monitoring scripts. Trigger power cycles from CI/CD pipelines. Build fleet management workflows that run on a schedule. Every command supports --json for machine-readable output.
 
The CLI covers roughly 47 commands across deployments, organization management, and authentication.

Interactive CLI

We also provide the CLI in interactive mode for engineers. Run the CLI with no arguments and you get a navigable menu for every operation. Arrow keys to browse, enter to select, guided prompts for each action.
 
The CLI is also available directly inside the Hydra browser app - same commands, same output. For teams evaluating the tool before a local install, access is available through the Private Beta program.

API v1.0: The Stable Foundation

The Hydra API v1.0 is the stable, versioned layer behind everything the CLI does. Now in General Availability, the API provides documented endpoints across device management, organization control, authentication, and system health - with a full OpenAPI spec published at hydrahost.com/api/redoc.
 
Every CLI command maps to an API endpoint. Teams building custom orchestration, programmatic provisioning pipelines, or event-driven automation via webhooks can build directly on the API. It is backwards-compatible, so integrations built today will not break tomorrow.

Getting Started

API v1.0 is in General Availability. Explore the full documentation and OpenAPI spec at hydrahost.com/api/redoc.
 
The Hydra CLI is in Private Beta. Request access through Hydra Support and your team can be running commands within minutes of installation.

Explore the API Documentation | Request CLI Beta Access