On-premise AI agents
On this page
Quick startYour dataYour chosen LLMOmnimatch stays on your infrastructureChoose how your deployment connectsRun Omnimatch on your infrastructure, choose the data each agent can use and decide where its LLM runs. Keep everything local or connect only the services your organisation approves.
Your data
- Uploaded documents
- OneDrive
- Google Drive
- Other company data
Your chosen LLM
- Run it locally
- Use your own provider
- Use Omnimatch models
Keep both sides local, or connect only the services your organisation approves.
Quick start
Use a Linux server or cloud VM with Docker Compose. Start free with one user and one agent. No product key required.
Run these commands on your server to download and start the public image. No registry login is needed:
mkdir omnimatch # Create the installation folder
cd omnimatch # Enter the folder
curl --fail -O https://omnimatch.ai/on-prem/compose.yaml # Download the Compose configuration
docker compose pull # Download the latest image
docker compose up -d # Start in the background
docker compose logs -f # Follow startup logs
Press Ctrl+C to stop following the logs. Omnimatch keeps running in the background.
Open your installation’s HTTPS address and follow the wizard to create your administrator account and connect your LLM. For a remote server, set up HTTPS access first. Use http://localhost:8787 only when your browser is on the Docker host.
Your data
Give each agent access to the information it needs to do its job. Upload documents held on your infrastructure, connect selected files and folders from OneDrive or Google Drive, or connect other approved company systems with our help. Your organisation owns the provider accounts and credentials.
Your chosen LLM
Run a tool-capable model on your own infrastructure, connect an external provider your organisation already uses, or choose from selected Omnimatch models. The model connection is shared across the installation.
A local model supports fully air-gapped operation. External providers and Omnimatch models require internet access, and model requests leave your infrastructure. Check the provider’s retention terms or the provider zero data retention status shown for an Omnimatch model.
Omnimatch stays on your infrastructure
The application, agent settings, documents, conversations and encrypted credentials remain in your persistent Omnimatch installation. Your team manages its HTTPS address, storage, backups, updates and network policy.
Choose how your deployment connects
Keep everything local
Use local documents, approved internal data and a local LLM. This creates a fully air-gapped deployment that can operate without internet access.
Explore the air-gapped solution
Add approved connections
Keep Omnimatch on your infrastructure while connecting selected document providers, company systems or an external LLM. Your organisation controls which connections are allowed.