Skip to main content

Nora

The self-hosted AI agent ops platform.
Deploy, monitor, and operate OpenClaw and Hermes runtimes from one operator surface — multi-runtime, Apache 2.0, and on infrastructure you control. Nora gives technical teams a single place to deploy supported agent runtimes into isolated environments, manage LLM provider keys, access chat, logs, terminal, and live filesystem workflows, connect integrations and channels, and monitor runtime activity, all from one dashboard. The core value is straightforward: if you care about infrastructure control, observability, repeatable operations, and a trustworthy operator surface, Nora gets you to a usable ops platform faster than building one. It is designed for serious operator use — individual builders can run it on a single machine, internal platform teams can standardize on it, and larger teams can keep full infrastructure control without giving up the open-source trust model. Nora operator dashboard — system overview, fleet health, and recent activity

Key capabilities

Deploy agent runtimes

Launch agent runtimes into isolated environments with configurable CPU and RAM plus target-specific storage behavior. Docker and Remote Docker retain disk_gb as requested metadata rather than enforcing a filesystem quota. Manage agent lifecycle — deploy, restart, and remove — from a single operator surface.

Manage LLM provider keys

Store API keys for Anthropic, OpenAI, Google, and other providers, encrypted at rest with AES-256-GCM. Sync keys to running runtimes on demand.

Monitor operations

Track agent health, queue state, logs, metrics, and runtime activity in real time. Inspect individual agents or get a system-wide overview from the dashboard.

Integrations and channels

Wire external tools and configure communication channels directly from the control plane. Manage the full integration surface from one place.

Who Nora is for

Nora is built for the teams that care about how agent infrastructure runs, not just that it runs.

Internal AI platform teams

Teams that need a repeatable, auditable way to deploy and operate agents across their organization without building a control plane from scratch.

Technical product teams

Teams that want infrastructure control and observability as first-class features while keeping the full stack open and inspectable.

Ops-minded operators

Operators running agent infrastructure who need lifecycle management, logs, terminal access, and runtime monitoring in one surface.

Service providers

Teams that want to host and operate Nora-based ops platforms for clients on infrastructure they control.

Positioning pillars

Nora is built around five properties that define what kind of product it is and who it is credible for.

Security built in

Key encryption, JWT auth, secret generation, and CORS controls are baked in — not bolted on. Nora is designed to feel trustworthy in real operator environments.

Easy to use

On local Docker installs, the fastest path to first proof is short: install, create an operator account, launch the built-in zero-key demo agent, and validate chat, logs, metrics, and terminal access. Kubernetes-only installs start with a configured model provider.

Expandable

OpenClaw is the best-supported runtime today and Hermes is the second supported family. Runtime contracts, provisioner backends, CLI commands, and MCP tools are kept in public source so contributors can extend the supported surface without relying on a private plugin system.

Self-hostable

The repo, install scripts, and Docker Compose flow are the primary trust path. Everything you need to run Nora is in the open repo. You own the infrastructure.

Enterprise-capable

Nora makes sense from single-host Docker evaluation through Kubernetes, private cloud, on-prem, operator-registered Remote Docker hosts, and AWS / Azure / GCP rollouts. Remote Docker and Proxmox unprivileged LXC placement are experimental and require validation on the exact infrastructure before production use. It is not a toy dashboard — it is designed to grow with operator requirements.

Apache 2.0 licensed

Self-host, modify, use commercially, host for clients, or build integrations on top of Nora. The open-source repo and self-hosted workflow are the center of the product, not a free tier.

Deployment footprint

Nora is designed to grow with your infrastructure requirements. Lead evaluation with a GA path; experimental and roadmap targets remain clearly labeled. Remote Docker is not a capacity scheduler or failover layer. Its connection badge is the stored result of a manual backend-only test, and operators select one exact remote:<id> target per agent. Review the Remote Docker guide before exposing a host or sharing it with a workspace.
NemoClaw on Proxmox remains blocked. Do not present Proxmox as GA or as a hardware-isolation path until the real-hardware release gate is complete.

Get started

Try the zero-key demo

Create an operator account and, when local Docker is enabled, deploy the deterministic demo agent before supplying an external provider key.

Quick start

Install Nora, create your operator account, launch the demo or add a provider, and validate your first agent runtime.

Self-hosting

Run Nora on your own infrastructure with the installer script or manual Docker Compose setup.

API reference

Explore the REST API for agents, providers, integrations, channels, workspaces, and monitoring.

Guides

Step-by-step walkthroughs for deploying agents, managing providers, wiring integrations, and more.

Contribute

Pick a focused issue, improve a runtime or backend contract, or help harden the operator experience.