Deterministic pipelines
YAML-defined DAG workflows with parallel step execution, dependency resolution, context propagation between steps, and automatic retry with configurable fallback strategies.
Orchestrate AI agents on
your own VPS and your own terms..
Your agents run on scattered machines with no unified way to dispatch, monitor, or stop them.
No visibility into what agents are doing in real-time. Failures are discovered hours later in logs.
Task dispatch, approvals, retries, escalations — all requires SSH-ing into machines.
Prompts and agent instructions are copy-pasted, never versioned. No skill reuse across agents.
Complex tasks require multi-step coordination. Without a workflow engine, you chain scripts and pray.
Security, networking, container isolation — bolted on later instead of built in from day one.
↓
YAML-defined DAG workflows with parallel step execution, dependency resolution, context propagation between steps, and automatic retry with configurable fallback strategies.
SKILL.md files synced natively to agent workspaces. Import from the marketplace, create your own, or version them in Git. Agents load skills automatically.
Cron-based repo scanning with multi-agent parallel analysis. Auto-generates structured proposals with risk assessment. Approve or let it execute.
Live task feeds via Convex real-time subscriptions. Workflow step-by-step progress, agent status, cost tracking per task. Two UIs: Dashboard and Mission Control.
Cortex runs on your own servers. Here is what powers it.
Your agents run on your own machines. Full control over CPU, memory, and storage. No shared infrastructure.
Zero-trust networking between your VPS and Cortex. No public ports. Encrypted WireGuard tunnels. Agents unreachable from the internet.
Each agent runs in its own isolated container with the OpenClaw runtime. WebSocket protocol v3 for bidirectional communication. Workspace-level file isolation.
Push-based architecture: Convex dispatches tasks to the gateway, which routes them to the right agent via OpenClaw. No polling. Real-time response streaming.
Serverless functions, real-time subscriptions, and transactional consistency. Dashboard and Mission Control get instant updates without refresh.
Configure agents, launch workflows, and monitor everything in real-time. Powered by Convex subscriptions.
Cortex is built on OpenClaw, an open-source AI agent runtime. Your agents, your infra, your data. No vendor lock-in.
No vendor lock-in.
Early adopters get direct access to the founding team. Your feedback drives what we build next.
Your feedback drives what we build next.
Production-ready agent orchestration before the market catches up. Skills, workflows, and infrastructure — ready now.
Ready now.
Deploy on Hetzner, AWS, or any Linux host. One Docker Compose file brings up OpenClaw agents, the WebSocket gateway, and the reverse proxy.
Tailscale mesh VPN connects your VPS to Cortex. Agents run in isolated Docker containers. No public ports exposed except the gateway.
Define agent roles, assign skills, create YAML workflows. Everything is managed from the Cortex dashboard — no SSH needed after initial setup.
Tasks flow through the gateway in real-time. Review proposals, approve workflows, track costs. Add agents as you grow — same infrastructure, more capacity.
All prices in EUR. No hidden fees. Cancel anytime.
For solo developers shipping with one agent.
For teams automating their development workflow.
For organizations that need full control at scale.
Start with a Hetzner VPS and a Docker Compose file. Full orchestration in minutes.
Free tier available. No credit card required.