3D in HydraOps: the model writes Three.js, a sandbox renders it

The 3D plugin has no mesh generator behind it. You describe an object, your model (the local one too) writes the Three.js code, and an isolated iframe renders it. Iterate in plain language, fix errors automatically, and export .glb for Unity, Unreal or Blender. Everything runs on your machine, offline.

Your agents now remember, without a vector database

Two small tools give HydraOps agents memory: remember keeps curated, dated notes in a Markdown file, and recall searches their own past conversations with SQLite full-text search. No embeddings, no extra service, and no agent can read another one.

One Shot: draw the idea, compile the prompt

The first mini-app in the HydraOps Complementos hub turns a diagram of nodes and arrows into a single self-contained prompt, compiled by your own configured LLM. With three worked examples: a game, a horror story and a mobile app.

Your agents now answer on Telegram and work on GitHub

HydraOps grows its first two connectors: a Telegram bot to run your agents from your phone, with push notifications for scheduled tasks, and a GitHub tool gated per agent. Both tokens live behind the credential firewall.

Your AI agents do not need your API keys

How the HydraOps credential firewall keeps provider keys out of the repository, the database and the agents’ reach. A ~130-line local proxy injects them at the network boundary.