# Introduction

Getting started with Knowns CLI - the knowledge management tool for developers

Knowns is a CLI-first knowledge and workflow layer that gives AI persistent memory of your project context. Document once, link everywhere.

**v0.20** brings native ONNX embedding (no sidecar needed), consolidated MCP tools, AI permission guards, audit trails, and structural knowledge retrieval.

## Latest Releases

- **v0.20.1**: Fix ONNX runtime lib resolution in sibling dirs
- **v0.20.0**: Native ONNX embedding, MCP tool consolidation (46 → 10 domain-grouped tools), AI permission model, MCP audit trail, structural knowledge retrieval, Windows native support
- **v0.19.x**: Semantic reference runtime, tree-sitter sidecar, design system reference
- **v0.18.x**: Code intelligence, code graph UI, workspace switching, chat UX, 3-layer memory, runtime adapter install

## How To Use These Docs

If you are new to Knowns, read the docs in this order:

1. Installation
2. Task Management
3. Workflow
4. CLI Reference
5. Feature-specific pages such as MCP, memory, code intelligence, or Web UI

## Why Knowns?

Modern development involves juggling tasks, documentation, and context across multiple tools. Knowns brings everything together in a simple, git-friendly format that works seamlessly with AI assistants.

### Key Benefits

- **Semantic Search**: AI-powered search using native ONNX embedding. Find by meaning, not just keywords
- **AI Chat**: Built-in project-aware chat inside Knowns, powered by OpenCode
- **Spec-Driven Development**: Write specs first with `/kn-spec`, link to tasks, auto-sync ACs
- **CLI-First**: Everything accessible from your terminal
- **AI-Ready**: Automatic context loading via `@doc/...`, `@task-...`, and `@template/...` references
- **Skills System**: `/kn-*` namespace for task, spec, plan, implement, verify workflows
- **Code Intelligence**: Dedicated code search, symbols, deps, and graph flows for code-aware AI work
- **Memory System**: Working, project, and global memory layers for durable AI context
- **AI Permission Model**: Capability-based permission guards for AI actions across MCP and runtime
- **MCP Audit Trail**: Structured audit events and aggregate tool usage statistics
- **Structural Retrieval**: Relation-aware knowledge retrieval across entities and graph edges
- **Dashboard + Web UI**: Faster local interface for Kanban, docs, memory, graph, and chat
- **Knowledge Graph**: Visualize relationships between tasks, docs, and memories
- **Template System**: Generate code with Handlebars templates linked to documentation
- **Import System**: Share docs and templates across projects from Git, npm, or local sources
- **Git-Friendly**: All data stored as Markdown in `.knowns/` folder

## Prerequisites

Knowns supports multiple install paths. Only the npm-based path requires **Node.js 20+**. See the [Installation Guide](https://knowns.sh/docs/installation) for platform-specific instructions.

## Quick Start

```
# Install
curl -fsSL https://knowns.sh/script/install | sh

# Initialize in your project
cd your-project
knowns init

# Create your first task
knowns task create "Setup project" -d "Initial setup" --ac "Dependencies installed"

# Open local UI + AI Chat
knowns browser --open

# (Optional) Auto-setup MCP for AI assistants
knowns mcp setup
```

## Core Concepts

### Tasks

Tasks are the primary work items in Knowns. Each task has:

- **Title**: Clear summary of what needs to be done
- **Description**: Context and requirements
- **Acceptance Criteria**: Testable outcomes
- **Status**: todo, in-progress, in-review, blocked, on-hold, done, urgent

### Documentation

Documentation lives in `.knowns/docs/` as Markdown files with frontmatter. Reference docs anywhere using `@doc/path` or `@docs/path` syntax (both work identically).

### Memory

Three-layer memory system for persistent AI context:

- **Working memory**: Session-scoped, ephemeral
- **Project memory**: Repo-specific patterns, decisions, conventions
- **Global memory**: Cross-project preferences and workflow rules

### Time Tracking

Built-in timers let you track time spent on tasks. Generate reports by date range, label, or project.

### Templates

Code generation with Handlebars (`.hbs`) templates powered by Go's `text/template` engine. Each template can link to documentation for context-aware generation.

```
# List templates
knowns template list

# Generate code
knowns template run component --name Button
```

## Next Steps

- [Installation Guide](https://knowns.sh/docs/installation) 
- [Task Management](https://knowns.sh/docs/tasks) 
- [Memory System](https://knowns.sh/docs/memory) 
- [Code Intelligence](https://knowns.sh/docs/code-intelligence) 
- [Semantic Search](https://knowns.sh/docs/semantic-search) 
- [Template System](https://knowns.sh/docs/templates) 
- [Import System](https://knowns.sh/docs/imports) 
- [CLI Reference](https://knowns.sh/docs/cli-reference) 
- [Philosophy](https://knowns.sh/docs/philosophy) 
- [Contributing](https://knowns.sh/docs/contributing) 
- [Roadmap](https://knowns.sh/docs/roadmap) 
- [License](https://knowns.sh/docs/license)

### On this page

Knowns v0.20.5

A new version is available