Self-hosted voice AI
in your own cloud
Run ThinnestAI inside your own AWS, GCP, Azure or on-prem VPC. Same platform, same flow editor, same real-time engine — with your data, your keys and your network boundary. Built for regulated BFSI, healthcare, insurance and government deployments.
When self-hosted is the right choice
Data residency
Patient records, borrower PII, government citizen data — some workloads simply can't leave your infrastructure. Self-hosted ThinnestAI keeps every call recording, transcript and log inside your VPC.
Regulated compliance
BFSI, healthcare, insurance and government teams often need auditor-grade isolation. Your data never touches a third-party SaaS — you control access, retention, and deletion end to end.
Integration depth
Connect voice agents to internal databases, mainframes, core banking systems or private APIs that aren't exposed to the public internet. Self-hosting puts the agent runtime on the same network as your systems of record.
Bring your own cloud
Deploy on AWS, Google Cloud, Azure or your own on-prem Kubernetes cluster. Same platform, your servers, your network, your keys.
What you get in self-hosted
The full ThinnestAI platform — identical to the hosted version — running inside your VPC.
Deployment targets
When self-hosted is NOT the right choice
Self-hosted adds real operational cost — your team runs the Kubernetes cluster, handles upgrades, owns incident response. Before choosing self-hosted, check that you actually need it:
- —If you just want the standard platform with good security, the hosted version with Mumbai region data residency is usually enough.
- —If you're a startup or mid-market team without dedicated infrastructure engineers, self-hosting will slow you down.
- —If your compliance requirements can be met with a signed DPA and data residency certification, hosted is faster and cheaper.
- —Self-hosted is the right fit when data residency, integration depth, or regulatory audit requirements genuinely force your hand — not when it just feels safer.
Frequently asked questions
Can I self-host ThinnestAI?
+
Yes. ThinnestAI supports self-hosted deployment in your own AWS, GCP or Azure VPC (or on-prem Kubernetes) for enterprise customers. You get the same platform, flow editor, real-time engine and integrations — running entirely inside your infrastructure, with your data staying in your VPC.
Why would I self-host a voice AI platform?
+
The two biggest reasons are data residency (compliance requirements that prohibit customer data leaving your infrastructure) and integration depth (direct access to internal databases, tools and systems that can't be exposed over the public internet). Regulated industries — BFSI, healthcare, insurance, government — typically need one or both. Some enterprises also self-host for latency reasons (running the agent runtime on the same network as their systems of record).
What's included in the self-hosted deployment?
+
The full platform. No-code flow editor, real-time voice engine, agent runtime, dashboards, observability, webhook dispatch, flow versioning, RBAC and SSO. You still use external LLM / STT / TTS providers via BYOK — either calling out over your own outbound network or routing through a private API gateway.
Do external provider calls leave my VPC?
+
That depends on how you configure it. If you BYOK OpenAI, Anthropic, Deepgram or similar, the agent runtime in your VPC makes outbound HTTPS calls to those providers — so voice data does transit over the public internet to the provider and back. For the strictest data residency requirements, you can route through a private API gateway or use fully-local providers (e.g., Sarvam via a private endpoint). We'll help you design the topology.
How long does a self-hosted deployment take?
+
For a standard VPC deployment on AWS / GCP / Azure, typically 2–4 weeks from contract signature to first production call. The long pole is usually internal security review and network access provisioning on your side — the actual platform deploys in hours via our Terraform modules.
How does pricing work for self-hosted?
+
Self-hosted deployments are enterprise contracts, priced per platform seat and volume rather than the standard ₹1.5/min. Contact sales for a custom quote — pricing depends on deployment scope, support level, the number of agents you run, and whether you need 24/7 dedicated support.
Do you support air-gapped deployments?
+
For fully air-gapped deployments (no outbound internet access at all), the LLM / STT / TTS providers have to run inside your network too — either self-hosted open-source models or on-prem inference from providers that support it. This is a custom engagement — talk to us about scope.
How do updates work?
+
ThinnestAI pushes platform updates via a private channel to your cluster on a cadence you control. You can pin versions, roll back, and test updates in a staging environment before production. Security patches are fast-tracked with your security team's notification.
Talk to our enterprise team
We'll walk you through the self-hosted architecture, compliance posture, and deployment timeline for your specific environment.