ORBIT
A self-hosted AI gateway connects local or hosted models to private files, databases, and agent tools through an OpenAI-compatible API.

ORBIT is a self-hosted AI gateway for teams building assistants over internal data, especially in regulated environments. It brings model access, data retrieval, and agent tools together on infrastructure the team controls.
- Questions across private data: Ask questions about PDFs, documents, and images, or combine results from SQL, MongoDB, and HTTP APIs in one answer. Follow-up questions can reuse retrieved datasets, and conversations can branch from any turn.
- Model choice: Use local or hosted models through one OpenAI-compatible API, with provider fallbacks and circuit breakers to handle service failures.
- Agent tools: Expose internal data and actions through an MCP tool server for Claude Desktop, Cursor, OpenClaw, or custom agents. ORBIT also supports communication between agents through A2A.
- Access and oversight: Manage access with Entra ID or Auth0 SSO and role-based permissions. An admin panel brings together health monitoring, sessions, API keys, usage controls, and audit logs.
- Voice assistants: Build full-duplex voice conversations, including assistants that answer questions about database content.
ORBIT is open source under Apache 2.0 and can run on-premises, in a private cloud, or in Docker. Air-gapped deployments use local models, keeping files and queries on your infrastructure. Hosted model providers require sending requests outside that environment.
Schmitech offers paid deployment, integration, training, and support from ORBIT's maintainers. It doesn't provide a hosted AI service or collect customer data; deployments remain under the customer's control.