Home / Deployment

Your bank’s own private copy.

Not a shared tenant. Each bank gets its own copy of Tuvel, with its own database, models, connectors and data handling.

  • A private copy per bank

    Its own database and a fixed outbound IP address to allowlist, inside the bank’s private VPC on Google Cloud, AWS or Azure, in the region the bank chooses.

    Available now
  • Model routing per agent

    Each agent runs on the model set for it, through one routing layer.

    Available now
  • Connectors per agent

    Your systems through OpenAPI or MCP, granted to one agent at a time.

    Available now
  • GCP, AWS or Azure, in your region

    The bank chooses the cloud and the region. Tuvel and its models run there, inside the bank’s own VPC.

    Available now
  • AI model calls inside India

    An open-weight model hosted in India keeps every AI model call inside India.

    Available now
  • Single sign-on with your identity provider

    Staff sign in through the bank’s own identity provider, with multi-factor authentication, passkeys and recovery codes.

    Available now
  • Reach internal systems without inbound ports

    An on-premises relay, so agents reach internal systems without a port opened to the internet.

    Per deployment

Models.

Each agent runs on the model set for it, through one routing layer that falls back when a key runs out and prices every call against the agent’s budget. Changing an agent’s model changes cost and speed, never its permissions.

Connectors.

Your systems connect through an OpenAPI description or an MCP server, granted to one agent at a time and revoked in one click, over mutual TLS, signed requests or OAuth. Calls leave from one fixed IP address you can allowlist, and an on-premises relay reaches internal systems without inbound ports.

Integrations

Data handling.

  • Consent first

    Accounts are read only after the customer agrees.

  • Masked identifiers

    Personal identifiers are masked before text reaches a model.

  • Credentials kept out

    PINs and one-time codes stay out of AI models.

  • Documents read, not stored

    Uploaded documents are read on the server; the original is not kept.

  • Encrypted

    Data is encrypted in transit and at rest.

  • Granted access

    Each agent sees only the data and tools it was granted.

Plan your deployment with us.

Bring your technology and security leads; we go through hosting, models, connectors and data handling.

Or write to demo@tuvel.co.