Not Diamond

Not Diamond is an intelligent model routing layer for coding agents. It predicts which AI model to use for each prompt, along with reasoning effort, to balance accuracy against cost. Rather than replacing an AI gateway, it integrates with existing gateway and harness infrastructure, so routing decisions are executed within the stack teams already use.

The router evaluates payload semantics, KV cache state, feedback signals, session outcomes, sub-agent architecture, compaction events, and prior turns against a model pool and cost objective. It supports multiple harnesses and providers, offers a savings and usage dashboard, and applies reinforcement learning to improve recommendations over time. Privacy-preserving routing restricts decisions to derived metadata and locally computed features. The platform is SOC-2 and ISO 27001 compliant, with custom SLAs and on-call support available.

Not Diamond is aimed at developers, small teams, and enterprise organizations running coding agent workloads. It is packaged as a pay-as-you-go plan charged per million tokens routed, with an Enterprise edition offering volume-based discounts, organization-wide analytics, admin controls, SSO with SAML, and privacy-preserving deployment. Access to the pay-as-you-go plan requires an application, and pilots can be arranged with the team.