OSCAR is OPSWAT's internal AI platform, built to replace fragmented chatbots and department-specific assistants with one unified agentic operating layer. Each department gets its own agent — its own personality, knowledge base, model tier, tools, and workflow permissions — reachable from web widgets, Teams, Slack, or email.
What it does
- LangGraph orchestrator routes user intent through a governed agent graph.
- LightRAG grounds every answer in retrieved enterprise knowledge.
- MCP tools let agents act on enterprise systems, not just talk about them.
- Multi-provider LLM gateway — swap Claude, GPT, or Gemini per agent with zero code changes.
- Logs retrievals, tool calls, and admin actions for a full audit trail.
- Fact-checking and human approval gate sensitive actions before they run.
Built for governance
- Department agents with configurable personalities, knowledge bases, tools, and model routing.
- Permissions and row-level data scoping enforced at the retrieval and tool layers.
- Human-in-the-loop approval for destructive or sensitive operations.
Why it mattersOSCAR turns AI from a collection of disconnected experiments into governed enterprise infrastructure.
Built with
PythonFastAPILangGraphLightRAGMCPpgvectorReactAzure