Nodal

Nodal is an evaluation platform that ensures AI agents provide accurate analytics by catching silent errors—confident but wrong answers that slip through without alerts. It starts with a context layer built through interviews with domain experts, defining metrics, canonical tables, mandatory filters, and potential data pitfalls. Teams use Nodal to verify agent answers against trusted dashboards, establish governed ground truth, and detect when data or model changes cause answer drift.

The platform offers three usage tiers: an open-source context layer (Apache-2.0) with agent skills and local evaluation, a hosted MCP for team sharing, and enterprise self-service analytics. Nodal never accesses your database directly—agents query through your own read-only connector while Nodal serves only context and definitions. The maker built it with Claude Code.