REBYTE.AI

Product Overview

Rebyte is built on a simple architecture: Agent Runtime + Code Agent + Skill. We believe this straightforward approach can solve almost every software task in the world. No complex customization needed.

Agent Runtime

At the foundation, every task runs inside an isolated runtime environment. This is the underlying layer that makes everything possible.

  • Isolated Execution - Each task runs in its own runtime, completely isolated from other tasks
  • Security Controls - Runtimes are protected with security measures to ensure safe execution
  • Network Policy - Enterprise users can specify custom network policies for their runtimes
  • Persistent Storage - Runtimes persist forever unless you explicitly delete them
  • Consistent Resources - Each runtime provides 4GB of memory and 2 CPUs, enough for any task

Code Agents

We currently support four code agents that run inside the VMs:

  • Claude Code - Anthropic's AI coding assistant
  • Gemini CLI - Google's AI coding tool
  • Codex - OpenAI's code generation model
  • Rebyte Code - Our own agent, a focused fork of OpenCode optimized for Rebyte's specific use cases

Each agent brings different strengths. You can choose the best one for your task or experiment to find your preference.

Skills

Skills extend what code agents can do. We support two types:

Public Skill Store

A curated collection of skills available to all users. Browse and use skills created by the community and the Rebyte team.

Private Organization Skills

Organizations can upload their own skills that are only available within their organization. This is useful for company-specific workflows, internal tools, or proprietary processes that shouldn't be shared publicly.

Learn More