A transparent index of nRouter’s published evidence, measurement methods, adoption artifacts, technical authorship, and the proof still being built with customers and independent reviewers.
Evidence policy
No unnamed logos. No universal savings number. No benchmark without a method.
Last reviewed October 5, 2026 · written by Suresh · technically reviewed by Rama Surasani.
A proof surface is more trustworthy when it distinguishes first-party material, independent validation, and evidence that needs customer permission.
nRouter is early and does not publish unnamed logos or invented outcomes. The design-partner program is open; customer stories will be published only after the customer approves the attribution, workload, baseline, and measured result.
The public leaderboard is a first-party model and routing surface, not an independent audit. We are publishing the measurement protocol before asking an outside party to reproduce it: fixed models, warm-up policy, region, concurrency, p50/p95/p99 latency, error rate, and provider route logs.
A cost claim is only useful when it is tied to the same prompts, token counts, models, routing policy, and provider invoice. The cost estimator is available for scenario planning; audited savings require a customer baseline and are not presented here as a universal percentage.
No public incident postmortem is claimed until there is a real incident record to review. The status page is the source of truth for availability; future postmortems will include timeline, detection, customer impact, routing decisions, and corrective actions.
The security page describes the request path, tenant isolation, credit safety, session policy, data residency, and vulnerability reporting. Procurement evidence and attestations remain separate from architectural claims and are linked through the trust and compliance surfaces.
Developer documentation names its authors. This proof page is owned by Suresh, Senior SEO Engineer, and technically reviewed by Rama Surasani, Founder & CEO of CloudAct Inc. Product claims should still be checked against the linked source and current product docs.
The official SDK repository and public package registries are inspectable adoption artifacts. They show source, examples, release history, install instructions, and package metadata; they are not represented as independent customer adoption numbers.
This page does not label first-party social accounts as independent reviews. Community references will be added with the original URL, publication date, author, and context after they exist and can be verified.
The nRouter YouTube channel is the canonical place for product walkthroughs and future benchmark demonstrations. Each video should identify the date, environment, model, region, and whether the result is first-party or independently reproduced.
Before publishing a performance or savings number, record the conditions that make the comparison meaningful.
Model IDs, provider route, fallback order, region, and routing policy.
Warm-up count, concurrency, prompt/response size, p50/p95/p99, errors, and retries.
Provider list price, token counts, platform fees, cache state, and invoice period.
Threat model, test corpus, false positives, redaction results, retention controls, and reviewer.
Each tutorial keeps the integration path concrete and names compatibility limits instead of implying that every feature behaves identically through a gateway.
Python and JavaScript chains, RAG, routing, retries, and cost metadata.
Open tutorialMulti-agent crews with fallback models, guardrails, and budget isolation.
Open tutorialTypeScript and Next.js integration over an OpenAI-compatible endpoint.
Open tutorialRoute supported Claude Code traffic through a compatible gateway endpoint.
Open tutorialUse supported chat models with a provider key and verify Cursor feature limits.
Open tutorialConfigure an OpenAI-compatible provider and validate model and tool behavior.
Open tutorialEvidence program
Share a benchmark, security review, or customer outcome with the team. We will publish the method and limitations with the result—or explain why the evidence is not ready.