DVARA

DVARA is an AI gateway that governs model calls, tool calls and agent-to-agent calls, and deploys anywhere. It keeps a tamper-evident audit trail, runs policy and personal-data checks on every call, and stops agents that loop. Routing and OpenAI compatibility come standard.

What it covers:

• LLM calls: one OpenAI-compatible endpoint in front of multiple providers and local models, with routing, fallback, budgets and per-team spend control.
• MCP tool calls: policy checks and audit on every tool call an agent makes.
• Agent-to-agent (A2A) calls: delegation limits and loop detection, so one agent cannot keep calling another without bound.
• Policy as code: rules are reviewed, tested and promoted like any other code.
• Audit: every decision is written to a tamper-evident, hash-chained log.

The DVARA LLM Gateway is open source under Apache 2.0 and free to run in production. MCP and A2A governance run without a licence key for evaluation; production use of those needs a licence.