Private AI
Private AI. Your agents, your environment.
CAIDen voice, SMS and chat agents that run in your cloud account or on your servers, on open-weight models.
Your conversation data stays inside your environment, and conversation content is not sent to a third-party AI provider.
Calls and texts travel through your telephony carrier (Twilio).
Inside your environment
What runs where you control it.
Every layer of the deployment, and the options that run it in your environment.
Language model
Runs in your cloud account and on your servers
Speech to text
Runs in your cloud account and on your servers
Text to speech
Runs in your cloud account and on your servers
Knowledge base
Runs in your cloud account and on your servers
Agent runtime (Hermes Agent)
Runs in your cloud account and on your servers
Conversation data
Runs in your cloud account and on your servers
Call and message media (LiveKit)
Runs in your cloud account and on your servers
Workflows (Inngest)
Runs in your cloud account and on your servers
Calls and texts travel through your telephony carrier (Twilio).
Integrations such as HubSpot, Salesforce, Calendly, Stripe or Postmark connect only if you choose, to your own accounts.
CAIDen staff have no access to your conversation data. Support access only with your permission.
Deployment
Where it runs.
In your cloud account
Installed in your own cloud account.
- AWS
- Google Cloud
- Azure
- Your own account
SOC 2 + ISO 27001 infrastructure, held by your cloud provider (AWS, Google Cloud or Azure).
Your conversation data stays inside your environment, and conversation content is not sent to a third-party AI provider.
Calls and texts travel through your telephony carrier (Twilio).
On your servers
Installed on your own servers, sized to the models and load.
Your conversation data stays inside your environment, and conversation content is not sent to a third-party AI provider.
Calls and texts travel through your telephony carrier (Twilio).
On a CAIDen box
Coming soonA dedicated CAIDen device for your office. Coming soon.
Models and knowledge
Open models, your documents.
Models
Open-weight models such as Qwen and Llama, served with vLLM, chosen per deployment.
Private knowledge base
Answers from your documents, with sources.
The knowledge base runs in your cloud account and on your servers.
Regulated work
Built for regulated work.
Designed to support HIPAA, TCPA and CASL requirements.
- Healthcare
- Finance
- Legal
- Government
Process
How it runs.
01
Discovery
We map your calls, texts and chats, the systems they touch and where the deployment will run.
02
Build
We build the agents and connect only the integrations you choose.
03
Test
Your team tests the agents on real scenarios before anything goes live.
04
Launch
The agents go live in your environment.
05
Operate
CAIDen operates the deployment after launch.
CAIDen staff have no access to your conversation data. Support access only with your permission.
