Skip to content
Omnimatch DocsOpen app

Air-gapped solution

On this pageDownload on a connected machineTransfer and startConfigure local servicesNeed an external LLM?

Run your agent entirely on your infrastructure with local data and a local LLM. This solution works without internet access.

Your infrastructure

Your data

  • Local documents
  • Internal company data
OmnimatchOn your infrastructure

Your local LLM

  • Runs on your infrastructure
  • Supports tool calling
  • No external provider

No internet connection required. No model requests leave your network.

Download on a connected machine

Pull the public image for the isolated server’s architecture, then save it as an archive. Use linux/arm64 instead for an ARM64 server.

docker pull --platform linux/amd64 ghcr.io/omnimatch/omnimatch-agents:onprem-stable
docker save -o omnimatch.tar ghcr.io/omnimatch/omnimatch-agents:onprem-stable
curl --fail -O https://omnimatch.ai/on-prem/compose.yaml

Transfer and start

Transfer both files to an omnimatch folder on the isolated server. Open a terminal in that folder, then import and start:

docker load -i omnimatch.tar
docker compose up -d --pull never
docker compose logs -f # Follow startup logs

Press Ctrl+C to stop following the logs. Omnimatch keeps running in the background.

Configure local services

Before disconnecting the server, install Docker and provision a tool-capable model with an endpoint the Omnimatch container can reach. Make sure any approved internal company data is reachable from the container too. In the setup wizard, choose your local model and enable only local connections.

Need an external LLM?

Selected Omnimatch models use provider zero data retention, meaning the model provider does not retain prompts or replies. This applies to model processing only and requires an internet connection, so it is not fully air-gapped. See Choosing a model for availability.