What deploys. One docker compose stack: the operating console (the attention queue + per-domain workspace), six governed domain agents (with the learning + discovery engine), and the OSS cores they act through. All images are public on GHCR — pull and deploy them yourself, no login. The runtimes and control plane underneath install separately (core install).
| Business system | Domain agent | Acts through |
|---|---|---|
| Revenue | revenue-agent — prospecting + next-best-action | Twenty CRM |
| Intelligence | intelligence-agent — growth · BI · market radar | Metabase |
| Finance | finance-agent — billing + books ops | books / billing |
| Customer Success | customer-success — support · guide · lifecycle | support desk |
| Content | content-agent — content engine | Postiz |
| Security & Compliance | security-compliance — KYC · controls | controls / evidence |
What you need. A Linux, macOS or Windows machine or server with administrator rights (root / sudo / local-admin) and at least 32 GB RAM — Twenty, Metabase, Postiz and their databases are heavy. Docker + Docker Compose v2. And an OpenAI-compatible LLM endpoint the host can reach — cloud, or a model on your own hardware (Ollama · local vLLM · NVIDIA PAIR). The app images are public, so no registry login is required. This is not a phone/tablet app and not a hosted SaaS you log into — you run the stack on hardware you control.
Deploy the core first. The apps assume rAIls are available. If you haven't installed them yet, start with core install — then come back here.
Questions or stuck? Join the ReDevOps Slack — #agentic-apps is the channel for the business suite (deploys, cores, the operating console). We answer there.
Prefer an app to a Compose file? rAIlsApps is a native desktop app — the control surface for the whole suite: Sidekick chat, Projects & Missions, approvals, evidence & status, the Connection Manager, and deploy / upgrade / health controls, in one window. It installs without Docker and either connects to a Runtime you already run or deploys one for you.
Package the experience; deploy the infrastructure as infrastructure. The desktop app is the UI. The runtimes, domain agents, databases and OpenBao/Vault keep running as host / container / cloud services behind a small, auditable host controller (redevopsd) that verifies signed image digests, provisions secrets, and does start / upgrade / rollback with receipts. The UI never receives a root shell or your plaintext secrets.
| Channel | Platform | Role | Status |
|---|---|---|---|
| Flathub / Flatpak | Linux desktop | Primary — one-click install + repository updates (candidate id io.redevops.AgenticApps) | Early access |
| .deb + apt | Ubuntu / Debian | Best for managed hosts and the redevopsd controller service | Early access |
| AppImage | Linux portable | No-install evaluation fallback | Early access |
| Signed MSI / MSIX | Windows | Desktop app + optional helper service (Docker/WSL2 only for the full stack) | Planned |
| Notarized DMG | macOS | Homebrew cask optional; heavier workloads deploy to a remote target | Planned |
| Container images / Helm | Server · VM · Kubernetes | The Runtime / data-plane deployment behind the desktop — self-host below | Available now |
Early access. The desktop packages are rolling out channel by channel — Flathub is the lead Linux channel. Until your platform's channel is live, the fastest way to run the complete suite today is the Docker self-host below. Ask in #agentic-apps for current early-access builds.
The whole suite ships as one Compose file in the public agentic-os repo (deploy/apps). Pick your OS — the steps are identical apart from how you install the prerequisites and generate secrets.
Prerequisites (Docker Engine + Compose v2 + Git):
sudo apt install -y docker.io docker-compose-v2 git openssl
sudo usermod -aG docker $USER # then log out/in so docker runs without sudo
Get the stack and bring it up:
git clone https://github.com/redevops-io/agentic-os cd agentic-os/deploy/apps cp .env.example .env # set the two required secrets + your tenant/LLM in .env sed -i "s|^TWENTY_APP_SECRET=.*|TWENTY_APP_SECRET=$(openssl rand -base64 32)|" .env sed -i "s|^POSTIZ_JWT_SECRET=.*|POSTIZ_JWT_SECRET=$(openssl rand -base64 32)|" .env # images are public on GHCR — no docker login needed docker compose up -d docker compose ps
Open http://localhost:8080 — the operating console.
Prerequisites (Docker Desktop + Homebrew):
brew install --cask docker # then launch Docker Desktop once
brew install git openssl
Get the stack, log in, and bring it up:
git clone https://github.com/redevops-io/agentic-os
cd agentic-os/deploy/apps
cp .env.example .env
sed -i '' "s|^TWENTY_APP_SECRET=.*|TWENTY_APP_SECRET=$(openssl rand -base64 32)|" .env
sed -i '' "s|^POSTIZ_JWT_SECRET=.*|POSTIZ_JWT_SECRET=$(openssl rand -base64 32)|" .env
# images are public on GHCR — no docker login needed
docker compose up -d
Open http://localhost:8080. In Docker Desktop → Settings → Resources, give Docker at least 24 GB of memory first (Apple silicon & Intel both fine).
Prerequisites (Docker Desktop with the WSL2 backend, PowerShell):
winget install -e --id Docker.DockerDesktop winget install -e --id Git.Git
Then run the deploy from inside your WSL2 distro (Ubuntu) so paths stay on the Linux filesystem:
git clone https://github.com/redevops-io/agentic-os
cd agentic-os/deploy/apps
cp .env.example .env
sed -i "s|^TWENTY_APP_SECRET=.*|TWENTY_APP_SECRET=$(openssl rand -base64 32)|" .env
sed -i "s|^POSTIZ_JWT_SECRET=.*|POSTIZ_JWT_SECRET=$(openssl rand -base64 32)|" .env
# images are public on GHCR — no docker login needed
docker compose up -d
Open http://localhost:8080. In Docker Desktop → Settings → Resources → WSL integration, give it at least 24 GB of memory.
:3010) — create an account, then a personal API key (Settings → APIs). Paste it into .env as TWENTY_API_KEY and run docker compose up -d revenue-agent.:3001) and Postiz (:4200) — finish their first-run setup; the intelligence / finance / content agents connect by service name automatically.connected: false on that agent's /health; nothing crashes.Safe by default. SHADOW_MODE=true keeps every agent in stage-for-approval mode — it drafts each action and a human approves it in the console; nothing reaches a real CRM, social account or customer until you approve. Flip it to false per tenant only once you've reviewed the behavior.
The suite is a Compose stack, so the cloud recipe is the same everywhere: provision one VM with ≥ 32 GB RAM, install Docker, then run the identical git clone → docker compose up from the self-host tab. Only the "create a VM" line changes per provider.
aws configure # r6i.xlarge = 4 vCPU / 32 GB. Open port 8080 in the security group. aws ec2 run-instances --image-id ami-0c7217cdde317cfec --instance-type r6i.xlarge \ --key-name YOUR_KEY --security-groups YOUR_SG ssh ubuntu@EC2_IP # then run the Linux self-host steps above (Docker + git clone + docker compose up)
Model: Amazon Bedrock via an OpenAI-compatible gateway — set REDEVOPS_LLM_BASE_URL in .env. Full AWS integration →
gcloud compute instances create redevops-apps \
--machine-type=e2-standard-8 --image-family=ubuntu-2404-lts \
--image-project=ubuntu-os-cloud --boot-disk-size=100GB
gcloud compute firewall-rules create allow-apps --allow tcp:8080
gcloud compute ssh redevops-apps
# then run the Linux self-host steps above
Model: Vertex AI via the Google Agent stack. Full Google integration → (e2-standard-8 = 8 vCPU / 32 GB.)
az vm create -g redevops-rg -n redevops-apps --image Ubuntu2404 \
--size Standard_D8s_v5 --generate-ssh-keys
az vm open-port -g redevops-rg -n redevops-apps --port 8080
ssh azureuser@VM_IP
# then run the Linux self-host steps above
Model: Azure OpenAI. Full Azure integration → (D8s_v5 = 8 vCPU / 32 GB.)
doctl compute droplet create redevops-apps --region nyc3 \
--size s-8vcpu-32gb --image ubuntu-24-04-x64 --ssh-keys YOUR_KEY_ID
ssh root@DROPLET_IP
# then run the Linux self-host steps above
Model: any OpenAI-compatible endpoint. Full DigitalOcean integration → (s-8vcpu-32gb Droplet.)
Reach it. The stack serves the console on port 8080. For a real hostname with automatic HTTPS, point a DNS record at the VM and set that domain in deploy/apps/Caddyfile (replace :80 with apps.example.com) — Caddy provisions a certificate on its own. Otherwise reach it at http://<vm-ip>:8080.
Once docker compose ps shows everything healthy:
for p in 8220 8221 8222 8230 8231 8232; do curl -fsS http://localhost:$p/health && echo # each domain agent done curl -fsS http://localhost:8080/health && echo # the console