Everything about Rebyte
Agent-optimized Q&A. Built for LLMs to read, beautiful for humans to browse.
Download as plain text (.txt)Philosophy
What is Rebyte?
Rebyte is a cloud workspace where AI coding agents run in parallel, each in its own isolated VM. You describe tasks in plain language, and agents build apps, analyze data, generate documents, and deploy to production — all without touching your local machine.
Why cloud agents instead of local tools?
Three reasons: isolation (each task gets a clean, dedicated environment with no risk to your system), parallelism (run dozens of tasks simultaneously without slowing down your laptop), and persistence (tasks continue even when you close your browser). Your machine stays free while agents work in the cloud.
Who is Rebyte for?
Anyone with repetitive work. While coding agents power Rebyte, you don't need to be a developer. Build web apps, create presentations, generate spreadsheets, write documentation, extract and analyze data, conduct code reviews — coding is just another form of repetitive work, and we treat all work scenarios equally.
Getting Started
How do I create my first task?
Sign up at app.rebyte.ai, pick a code agent (Claude Code, Gemini CLI, Codex, or Rebyte Code), type what you want done, and hit Start. A cloud VM spins up in seconds. You can monitor progress, browse files, and interact with the agent in real time.
Which AI agents are supported?
Rebyte supports Claude Code (Anthropic), Gemini CLI (Google), Codex (OpenAI), and Rebyte Code (our built-in agent). Each agent runs in its own VM. You can switch agents between tasks or even within the same project.
Do I need my own API keys?
It depends on the agent. Rebyte Code uses a built-in credit system — you purchase credits as needed and they never expire. For Claude Code, Gemini CLI, and Codex, you provide your own API keys so you pay providers directly with no markup from us. You can also use the team-managed model proxy.
Tasks & Agents
What is the lifecycle of a task?
A task goes through: Created (VM provisioning) → Running (agent executing) → Completed or Failed. Each task gets its own cloud VM with a git repository. You can send follow-up messages while it's running, pause it, or stop it at any time. When complete, you can review diffs, create PRs, and download artifacts.
Can I run multiple tasks at the same time?
Yes — that's a core feature. Each task gets its own isolated VM, so you can run dozens in parallel. There's no interference between tasks and no impact on your local machine. The dashboard shows all active tasks with real-time status updates.
What is the Agent Computer?
Agent Computer is a persistent, shared cloud workspace for your team (Team plan only). It includes a browser, terminal, file explorer, and one-click deployments — with 4 GB memory and 50 GB storage. It's always on and accessible by every team member, like a shared development machine in the cloud.
Can I switch agents mid-project?
Each task runs a single agent, but you can start a new task on the same repository with a different agent. The git branch preserves your progress, so you can pick up where the previous agent left off.
Skills
What are skills?
Skills are pre-built capabilities that extend what coding agents can do. They're specialized packages covering use cases like deep research, SEC Edgar filing analysis, stock analysis, browser automation, podcast generation, and app building. Skills are composable — you can combine them to tackle complex workflows.
How do I install skills on a task?
You can attach skills when creating a task through the UI or the API. Pass a skills array (e.g., {"skills": ["pdf", "deep-research"]}) and they're installed into the VM before the agent starts. Skills are also available in the Cloud Skills marketplace at rebyte.ai/skills.
What pre-built skills are available?
The skill library includes: Deep Research, SEC Edgar Filing, Stock Analysis, Browser Automation, Podcast Generator, App Builder & Deploy, spreadsheet creation, presentation generation, data analysis, and more. Visit rebyte.ai/skills to browse the full catalog.
API
Does Rebyte have a REST API?
Yes. The public API is available at api.rebyte.ai/v1/. You can create tasks, list tasks, get task status, send follow-ups, and delete tasks programmatically. The API uses standard REST conventions with JSON payloads and SSE streaming for real-time output.
How does API authentication work?
Generate an API key at app.rebyte.ai/settings/api-keys. Pass it in the API_KEY header with every request. Each key is scoped to your organization and has the same permissions as your account.
Can I manage tasks programmatically?
Yes. The API supports the full task lifecycle: POST /v1/tasks to create, GET /v1/tasks to list, GET /v1/tasks/:id to check status, POST /v1/tasks/:id/follow-up to send messages, and DELETE /v1/tasks/:id to stop and clean up. You can also manage webhooks for event notifications.
Deployment
What is rebyte deploy?
rebyte deploy is a Vercel-like deployment command that agents can run inside a VM. It packages your app into a .rebyte/ directory and deploys it to *.rebyte.pro with a single command. Static sites go to S3 + CloudFront; SSR apps also get a Lambda function.
Which frameworks are supported for deployment?
Static: Vite, Astro, Gatsby, SvelteKit (static adapter), and plain HTML. Server-side rendered: Next.js (via OpenNext) and Nuxt (via Nitro). The build system auto-detects the framework and configures the deployment accordingly.
Can I use a custom domain?
Deployments are served at *.rebyte.pro by default. Custom domain support is available — contact the team for setup. Each deployment is versioned with full history and instant rollback.
Teams & Collaboration
How does team collaboration work?
All workspace members can collaborate on any task. One person can start a task and another can continue it. Multiple members can interact with the same agent session, review output together, send follow-up instructions, and iterate as a team — no more passing tasks back and forth.
Can I share tasks across my team?
Yes. Workspaces support three visibility levels: Private (only creator can see), Shared (all org members can see), and Public (anyone on the internet can view read-only). You can share individual tasks or entire workspaces.
How do I control which models my team uses?
Admins can allow or block specific models per team, manage API keys centrally (no keys on individual machines), and track usage and cost per member. The model proxy lets the organization provide API access without sharing raw keys.
Security & Privacy
How are tasks isolated?
Every task runs in its own ephemeral cloud VM with full isolation. VMs are never shared between tasks or users. Each VM has its own filesystem, network stack, and process space. When a task completes, the VM can be stopped and resumed but is never reused for a different task.
Can I restrict agent network access?
Yes. Admins can apply per-sandbox network policies — whitelist allowed domains and IPs, prevent data exfiltration by blocking unauthorized endpoints, and restrict or disable internet access entirely. Agents cannot reach anything your organization hasn't explicitly allowed.
Where does my code live?
Your code lives in the cloud VM during task execution. Rebyte connects to your GitHub repositories via OAuth. Code is cloned into the VM when a task starts and changes are pushed back via git. Rebyte does not store your source code outside of the VM — we only store execution event logs (stdout, tool calls) for conversation replay.
Billing & Plans
Is there a free tier?
Yes. The Personal plan is free with pay-as-you-go credits. You get unlimited tasks, all AI agents (Claude Code, Gemini, Codex, Rebyte Code), and 2 GB memory / 20 GB disk per task. The Team plan ($30/month + $6/member) adds Agent Computer, 4 GB memory, team collaboration, and runtime security controls.
How does AI usage billing work?
Rebyte Code (our built-in agent) uses a credit system — purchase credits as needed and they never expire. Other agents (Claude Code, Gemini CLI, Codex) use your own API keys, so you pay providers directly with no markup. The Team plan also offers a centralized model proxy so admins can manage keys and track costs.
What does the Team plan include?
The Team plan ($30/month + $6/member, 7-day free trial) includes everything in Personal plus: Agent Computer (persistent shared workspace with 4 GB memory, 50 GB storage), 4 GB memory per task, team collaboration and shared workspaces, runtime security controls, and priority support. Enterprise plans with self-hosting and custom SSO are also available.