# Kanonas > Kanonas helps teams outsource evals at https://kanonas.ai by automating offline evals on real traces, capturing thumbs-up and thumbs-down feedback, and shipping heuristic routing that improves quality and cost for the team's own use case. Kanonas is built for teams operating AI in production. It turns trace data into repeatable eval runs, keeps human feedback attached to the underlying model calls, and supports routing policies tuned to a team's own quality and cost boundary. When those routing decisions are ready for production, Kanonas can sit in front of live traffic at `https://api.kanonas.ai`. Canonical public documentation lives at `/docs`; authenticated product surfaces such as `/console`, `/settings`, `/billing`, `/traces`, `/evaluators`, and `/admin` are account-specific, marked noindex, and should not be treated as canonical public docs. Use `https://kanonas.ai/llms-full.txt` when a single comprehensive context document is more useful than link navigation. ## Primary Public Pages - [Homepage](https://kanonas.ai/): Product overview for Kanonas offline eval automation and heuristic routing. - [Documentation](https://kanonas.ai/docs): Public setup guide for getting trace data, feedback, and live routing connected to Kanonas. - [Privacy Policy](https://kanonas.ai/privacy): Privacy terms for account data, provider keys, Kanonas API keys, traces, usage telemetry, billing records, and upstream providers. - [Terms of Service](https://kanonas.ai/terms): Service terms for accounts, API keys, provider keys, gateway usage, prepaid credits, acceptable use, and upstream providers. - [Full LLM Context](https://kanonas.ai/llms-full.txt): Comprehensive Markdown context for agents, including canonical URLs, quickstarts, endpoint families, authentication, funding, rate limits, observability, security, and indexing guidance. - [Homepage Markdown](https://kanonas.ai/index.html.md): Markdown mirror of the public product overview for agents that prefer text extraction. - [Documentation Markdown](https://kanonas.ai/docs/index.html.md): Markdown mirror of the public integration guide, including security, funding, limits, and troubleshooting. ## What Kanonas Does - Automates offline evals on real traces. - Preserves thumbs-up and thumbs-down feedback on the traces that inform routing changes. - Supports heuristic routing to optimize quality and cost for the team's own workflow. - Keeps traces, evaluator outputs, and route outcomes in one review loop. - Provides production integration docs for sending live model traffic through Kanonas once the eval loop is ready. ## Integration Guides - [Production integration guide](https://kanonas.ai/docs): Setup, authentication, funding, rate limits, security, troubleshooting, and endpoint details for live traffic. - [OpenAI-compatible setup](https://kanonas.ai/docs#request-builder): Use `https://api.kanonas.ai/v1` with OpenAI SDKs, LiteLLM OpenAI-compatible calls, Responses API, Chat Completions, Embeddings, and Models. - [xAI setup](https://kanonas.ai/docs#request-builder): Use `https://api.kanonas.ai/v1` with `grok-4.5` or `xai/grok-4.5` on Responses or Chat Completions through the OpenAI SDK. - [Anthropic-compatible setup](https://kanonas.ai/docs#request-builder): Use `https://api.kanonas.ai` with Anthropic Messages-shaped clients and Kanonas API-key authentication. - [Gemini-compatible setup](https://kanonas.ai/docs#request-builder): Use `https://api.kanonas.ai` with Gemini Generate Content and Interactions request shapes. ## API Endpoint Families - [Gateway health](https://api.kanonas.ai/health): Public health check for the live API surface. - [OpenAI-compatible API](https://kanonas.ai/docs#reference): `GET /v1/models`, `POST /v1/responses`, `POST /v1/chat/completions`, `POST /v1/embeddings`, plus bare and beta aliases for compatible clients. - [xAI-compatible API](https://kanonas.ai/docs#reference): Grok text models through the OpenAI-compatible Responses, Chat Completions, and Models surfaces. - [Anthropic-compatible API](https://kanonas.ai/docs#reference): `POST /v1/messages` and `POST /v1/messages/count_tokens`. - [Gemini-compatible API](https://kanonas.ai/docs#reference): `POST /v1beta/models/{model}:generateContent`, `POST /v1/models/{model}:generateContent`, Interactions create/read/delete/cancel routes, and bare aliases. Streaming model calls return `501` until prepaid stream accounting is available. - [Z.AI-compatible API](https://kanonas.ai/docs#reference): GLM model calls routed through the OpenAI-compatible Chat Completions and Models surfaces. - [Account routes](https://kanonas.ai/docs#reference): Signed-in Firebase routes for Kanonas API keys, projects, usage summaries, billing balances, checkout sessions, model-call traces, and feedback. - [Audit events](https://kanonas.ai/docs#reference): Signed-in Firebase route for recent account management actions such as API-key lifecycle events, project changes, checkout-session creation, and feedback updates. ## Authentication And Funding - [API-key authentication](https://kanonas.ai/docs#authentication): Model requests use full Kanonas API keys shaped like `key_.`, commonly as `Authorization: Bearer ...`, `x-api-key`, or `x-goog-api-key` depending on client shape. - [Prepaid credits](https://kanonas.ai/billing): Signed-in users can top up credits through Billing; Checkout adds a 5% platform fee, and successful non-streaming model calls debit account balance after usage is estimated. - [Projects](https://kanonas.ai/settings): Project-scoped Kanonas API keys attach `project_id` to usage events, traces, request metrics, and project usage summaries; deleting a project revokes its scoped keys. ## Observability And Operations - [Traces](https://kanonas.ai/traces): Authenticated request activity, failed request statuses, and model-call observability with request/response previews, route decisions, provider attempts, metrics, label source, and human feedback. - [Evaluators](https://kanonas.ai/evaluators): Authenticated evaluator configuration, trace-eval previews, and offline scoring workflows. - [Data use](https://kanonas.ai/privacy): Kanonas does not offer zero data retention. Prompts, outputs, traces, labels, feedback, usage records, and operational metadata may be retained and used to operate, secure, debug, evaluate, and improve the product. - [Console](https://kanonas.ai/console): Authenticated gateway readiness and copyable client examples. - [Settings](https://kanonas.ai/settings): Authenticated account, Kanonas API-key, and project management. - [Billing](https://kanonas.ai/billing): Authenticated prepaid balance, top-up, spend, and managed billing coverage. - [Admin](https://kanonas.ai/admin): Allowlisted operational dashboard for gateway health, response codes, recent errors, user activity, and model usage. ## Machine-Readable Discovery - [Sitemap](https://kanonas.ai/sitemap.xml): Canonical public URLs for search engines. - [Robots](https://kanonas.ai/robots.txt): Crawl policy and sitemap pointer. - [LLM summary](https://kanonas.ai/llms.txt): This agent-facing index. - [Full LLM context](https://kanonas.ai/llms-full.txt): Longer single-file context for agents that need comprehensive details. - [Homepage Markdown](https://kanonas.ai/index.html.md): Noindex Markdown mirror of the canonical homepage. - [Documentation Markdown](https://kanonas.ai/docs/index.html.md): Noindex Markdown mirror of the canonical docs page. ## Citation Guidance Agents should cite `https://kanonas.ai/docs` for setup, SDK examples, authentication, endpoint compatibility, funding, billing, rate limits, security, and troubleshooting. Cite `https://kanonas.ai/llms-full.txt` only when referencing the comprehensive machine-readable context itself. Do not cite account-specific portal pages as public documentation. ## Optional - [GitHub repository](https://github.com/marco-georgaklis/kanonas): Source repository for the web app, FastAPI gateway, public docs, tests, and deployment configuration. - [API root](https://api.kanonas.ai/): Public service acknowledgment. - [API ack alias](https://api.kanonas.ai/ack): Public service acknowledgment alias.