AcruxCore - 2026 Pricing, Features, Reviews & Alternatives

AcruxCore is an LLM-ops platform designed for engineering teams building and operating LLM-powered features in production. Deployed as a cloud-hosted service (no credit card required) or as a fully self-hosted stack running against a team's own database and provider keys, it sits between an application and model providers to centralize prompt management, request routing, observability, and quality evaluation under a single control plane. The platform is open source under the Apache License 2.0, with no features gated behind a paid tier.

The platform is organized into five composable modules that can be adopted independently or together. Prompt management stores versioned, templated message sets with named aliases (e.g., production, staging), allowing alias pointers to be moved between versions without redeploying the application. The gateway exposes a single OpenAI-compatible endpoint in front of OpenAI, Anthropic, and Gemini, normalizing differing request formats while supporting routing, cost accounting, exact-match response caching with configurable TTL, virtual keys, and per-key or per-team budgets. Tracing records every gateway call as a structured trace with spans covering model, token counts, latency, and cost, with no additional instrumentation required; custom spans can also be reported. The versioned tool catalog stores callable functions alongside prompts and passes them to models in the correct tool-calling format for OpenAI, Anthropic, or Gemini. The evaluation module builds datasets from real feedback, runs experiments across a prompt-version-by-model grid using a configurable judge model, and produces per-cell quality scores, regression deltas against a production baseline, a leaderboard, and an advisory winner.

AcruxCore targets engineering teams across any industry that need structured control over LLM prompts, costs, and quality in production. First-class async SDKs are available for TypeScript (Node 18+) and Python (3.9+), alongside a plain REST API, making adoption straightforward for teams already using OpenAI client libraries, which require only a base URL change to route through the gateway.