Run AI agents
like infrastructure.
Every project gets its own isolated environment with AI agents inside. Claude Code, Codex — running in containers on your Mac or Linux machine, sharing a filesystem, orchestrated from your terminal. Fully open source.
Isolation & Security
Let agents run loose. Safely.
Every vibespace is a sandboxed runtime. Agents get full root access inside their container — but can never escape it.
Container-per-agent
Each agent gets its own isolated container inside a VM. It can install packages, run servers, modify files — all without touching your host.
Stateful environments
Vibespaces persist across restarts. Code, dependencies, shell history, running processes — everything survives a reboot.
Bring your own keys
Your API keys go straight to the provider. No proxy, no middleman, no vibespace account required.
Resource limits
Set CPU and memory caps per agent. Kubernetes schedules them across your hardware. Pack dozens on a single machine.
One binary, no dependencies
Everything lives in ~/.vibespace/. No system packages modified, no Docker required, no sign-up. Just install and go.
Fully open source
Apache 2.0 licensed. Audit every line, fork it, run it air-gapped. No black boxes, no trust required.
Multi-agent orchestration
Multiple agents, one project.
Add Claude Code, Codex, or both to the same project. Share the filesystem for collaboration, or enable worktree mode so each agent gets its own branch for parallel development. Use @mentions to direct tasks to specific agents.
Agents
Pick your agents.
Claude Code, Codex, and more on the way. Same isolation, same commands. Mix them in the same project.
Claude Code
Anthropic
Full filesystem access, configurable permissions, tool control. Add more agents whenever you need them.
Codex
OpenAI
Same environment, same commands. Reasoning effort control, separate tool permissions.
OpenClaw
Coming soon
Open-source personal AI agent. Connects to messaging platforms, automates workflows, runs 24/7 in its own container.
OpenCode
Coming soon
Open-source terminal coding agent with LSP integration. Bring your own model — supports 75+ providers.
Remote Mode
Self-host your agent cluster.
Run vibespace on a VPS, connect from anywhere via WireGuard. Spin up 15 agents on a $10/month server. Your laptop stays cool.
vibespace remote connect <token> Scriptability
Built for automation.
JSON output, meaningful exit codes, headless mode. Drop vibespace into scripts, CI pipelines, or Makefiles.
--json Machine-readable output
{
"success": true,
"data": [
{"name": "backend", "status": "running", "agents": 2},
{"name": "frontend", "status": "stopped", "agents": 1}
],
"error": null,
"meta": {"count": 2}
} $? Meaningful exit codes
Headless mode
Send a task, get the result, exit. No interactive session needed.
Plain output
Use --plain for minimal output without colors or formatting.
CI/CD ready
Works in GitHub Actions, shell scripts, Makefiles. See the cookbook.
Get started
Install the CLI
One command to install. Pick your platform.
8GB RAM recommended · all install options
Agent skill for vibespace
npx skills add vibespacehq/vibespace-skill Don't want to self-host?
We're building a managed version. Drop your email and we'll let you know when it's ready.