Introduction

Knoku is an AI assistant for documentation sites. Unlike most alternatives, Knoku does not use embeddings or a vector database. It reads your docs the way a human would — as a hierarchical tree of pages, sections, and headings — and navigates that structure to answer questions.

This approach produces more accurate source citations, more predictable answers on technical content, and lets a single project serve multiple domains (marketing site, docs subdomain, locale subdomains) without re-indexing.

Instead of sending users to a search page or making them scan long guides, Knoku lets them ask questions in plain language and get an answer grounded in your documentation, with links back to the exact sections it used.

What Knoku Does

Knoku connects three parts of the documentation workflow:

• Sync your docs with the Knoku CLI.
• Answer visitor questions through the embedded widget.
• Track what users ask from the dashboard.

The goal is simple: help people find the right answer faster, while giving your team visibility into which parts of the documentation are useful, missing, or hard to understand.

How It Works

Knoku uses tree-based retrieval instead of vector similarity search.

When you push docs, Knoku parses each Markdown, MDX, or reStructuredText file and builds a tree from the heading hierarchy — pages become branches, sections become nodes. The original source is preserved alongside the tree so citations always point to a specific section.

When a visitor asks a question, Knoku traverses the tree to select the relevant sub-trees structurally, generates an answer grounded only in those sections, and streams the response back with links to the sources it used.

Because retrieval is structural rather than statistical, answers stay close to how your docs are actually organized, and every citation is traceable to a specific section. If a topic is not covered, the assistant will say so instead of guessing.

Main Components

CLI

The CLI is used by documentation maintainers. It detects supported documentation frameworks, stores project configuration, and syncs changed files.

The main CLI actions are init, push, status, and doctor. The Get Started guide shows the exact command format.

Widget

The widget is the chat interface that runs on your documentation site. It is loaded with a script tag or through the npm package, then renders inside a Shadow DOM so it does not interfere with the host site’s styles.

It supports streamed answers, source references, automatic light/dark mode, custom accent colors, suggested questions, and visitor identity metadata.

Dashboard

The dashboard is where your team configures and observes Knoku. Widget appearance and behavior themselves are configured in the embed code, while the dashboard handles everything around them.

It is organized around five areas:

• Projects — create one project per product or per independent docs surface. Each project has its own indexed docs and analytics.
• Allowed Domains — list every host the widget is allowed to load on. A single project can serve multiple domains (e.g. example.com, docs.example.com, locale subdomains, staging) and they share indexed docs, analytics, and plan limits.
• API keys & access — create keys for the CLI and for CI workflows, and manage which teammates can sync, configure, or read analytics.
• Analytics — see which questions visitors ask, which answers resolved the question, which sections of your docs are cited most, and where users hit topics your docs do not cover.
• Billing & usage — track monthly message usage against your plan limit and manage the subscription.

Supported Documentation Sites

Knoku is designed to work with common documentation platforms, static site generators, and custom documentation setups.

Supported documentation sources include:

• Docusaurus
• VitePress
• Nextra
• MkDocs
• Mintlify
• Sphinx
• Read the Docs (Sphinx/MkDocs)
• Custom documentation sites

For custom setups, Knoku can sync Markdown, MDX, or reStructuredText content when you provide the docs directory and project configuration manually.

When To Use Knoku

Knoku is useful when your documentation is large enough that users often need help finding the right page, section, option, or workflow.

It works well for:

• SaaS product documentation
• API documentation
• Developer guides
• Open source project docs
• Internal technical documentation

Next Step

Start with Get Started to create a project, sync your first docs, and install the widget — the free plan is enough to try the full flow on a real site.

See pricing  for plan limits and message allowances.