Open-source infrastructure for deploying, hosting, and managing fleets of AI agents across your infrastructure.

Website · Docs · Getting Started · Issues

OfficeOS is the infrastructure layer for running AI agents in production. Self-host the stack once, then launch agents on demand with their own virtual environment, attached browser, tools, persistent memory, credential access, structured logs, and isolated workspace.

The dashboard is the control plane for deploying agents, managing their capabilities, observing every turn, and operating them across teams and environments.

Self-host the full stack with Docker Compose or Kubernetes.

Preview

Quickstart	Integrations	Launch setup

Highlights

• One control plane — deploy, host, configure, and observe agents from the dashboard
• Fast agent launch — go from a configured control plane to new isolated agents in seconds
• Fleet scale — operate one agent or hundreds across your infrastructure
• Virtual environments — every agent gets an isolated workspace with filesystem and shell access
• Attached browsers — every agent has browser capabilities for web workflows and automation
• Managed tools and MCP — connect agents to built-in tools and MCP servers for external integrations
• Persistent memory — store agent memory, conversations, and operational context centrally
• Credential management — give agents scoped access to provider keys and integration secrets
• Structured logs — inspect message, tool call, tool result, and agent output timelines
• Model-agnostic — bring your own LLM keys for Anthropic, OpenAI, Google, xAI, and compatible providers

Quick Start

For local development, the root .env file is the source of truth for runtime config and secrets. Copy the template once, edit the provider keys, then start the development infrastructure from the repo root.

git clone https://github.com/HarKro753/EnterpriseAgentOs.git
cd EnterpriseAgentOs

cp .env.example .env
# Edit .env: add at least one LLM provider key

docker build -t harkro123/eaos-pod-executor:latest packages/pod-executor
docker compose -f docker-compose.infra.yml up -d

Kubernetes

OfficeOS is built for self-hosters who want to run agent infrastructure on their own cloud. Kubernetes manifests are available under k8s/ for production and staging deployments.

kubectl apply -f k8s/prod/

You can also install the core control plane with Helm:

cp k8s/helm/examples/values.local.example.yaml values.local.yaml
# Edit values.local.yaml with your database, public URLs, and provider keys.

helm upgrade --install eaos ./k8s/helm \
  --namespace eaos \
  --create-namespace \
  -f values.local.yaml

Development

Run infrastructure in Docker and run product code on the host for fast rebuilds. The backend creates one pod-executor container per agent and binds it to a random localhost port, so local dotnet run can call agent runtimes directly.

# Infra: Postgres, Redis, MinIO, channels, browser controller, browser node
docker compose -f docker-compose.infra.yml up -d
# Rebuild this when packages/pod-executor changes
docker build -t harkro123/eaos-pod-executor:latest packages/pod-executor

# Backend
cd apps/backend && dotnet run --project src/EnterpriseAgentOs.Api

# Dashboard
cd apps/dashboard && bun dev