Deployment and models
It lives where your architecture requires.
SaaS, your cloud, appliance or air-gapped, with the right model for each agent.
Four ways to run it
SaaS
We run it for you, under the controls your security team sets.
- Data
- In the environment we run for you
- Models
- The providers you choose
Your cloud
On Kubernetes, with OpenShift and AKS presets.
- Data
- Your database and storage, in your account
- Models
- Any provider, or a local LLM in your cloud
Appliance
A dedicated box on your premises.
- Data
- On the box, on your site
- Models
- A local LLM on the box, or a provider you allow
Air-gapped
Licensed offline, imported without a registry, no telemetry.
- Data
- Inside your enclave
- Models
- Local LLMs, inside the enclave
You choose which agent uses which model.
Any model, per agent, including local LLMs
Outside your boundary, through the egress proxy
- Anthropic
- OpenAI
- Gemini API
- Gemini Enterprise Agent Platform
- Microsoft Foundry
- Amazon Bedrock
Inside your boundary
- Ollama
- vLLM
- LM Studio
- llama.cpp
- SGLang
- TGI
Ollama, or any OpenAI-compatible server
Agents your teams already use run under the same controls.
- Avispa native agent
- Claude Code
- Codex
- GitHub Copilot CLI
- Gemini CLI
- Scripts and webhooks
In your cloud, on an appliance or air-gapped, Avispa sends no telemetry.