Documentation

Everything you need to get started with knktx. Install the CLI, connect your agents, and start shipping with persistent context.

Quick Start

Get up and running in under a minute. You need Python 3.10+ and at least one supported AI agent installed.

1. Open the app and grab your API key

> Visit app.knktx.dev → copy your API key from the welcome screen

2. Initialize in your project

> uvx knktx init

Detects your installed agents, configures MCP connections, and injects your rules and workflows.

3. Keep everything in sync

> uvx knktx update

Pull the latest rules, workflows, skills, plugins, and MCP configs from knktx into your local agent files.

Key Concepts

knktx organizes your AI agent context into 10 entity types. Every entity is fully manageable via MCP tools and the web app.

Projects: Top-level containers that group all context for a codebase or initiative. Each project has its own boards, documents, and settings.
Boards: Kanban-style task boards within a project. Agents move cards through columns (inbox → in_progress → done → merged) as they work.
Cards: Individual tasks on a board. Cards track branch names, PR URLs, assignees, and comments. Agents create, update, and move them via MCP.
Documents: Persistent project documents — handoffs, plans, project descriptions. Agents read these at session start to understand context.
Workflows: Step-by-step playbooks for specific task types (e.g., PR Planning, Bug Investigation). Agents follow these automatically.
Rules: Behavioral constraints agents always follow — prohibitions, obligations, and preferences. Rules can be global or scoped to a specific project. Injected into agent instruction files.
Skills: Reusable multi-file instruction sets agents invoke on demand. Skills support folder structures with multiple files, versioning, and can be shared via the public gallery.
Notes: Ideas, decisions, and observations that track their own implementation status. Link notes to cards for automatic progress tracking.
Plugins: Per-agent plugin backup and restore scripts. Store your full plugin configuration so any machine can be bootstrapped instantly.
MCP Configs: Agent-specific MCP server configurations. Manage which MCP servers each agent connects to, synced across all machines.

MCP Agent Setup

knktx works with any agent that speaks Model Context Protocol. The CLI auto-detects and configures your installed agents.

Claude Code

Auto-configured by knktx init

Gemini CLI

Auto-configured by knktx init

Codex CLI

Auto-configured by knktx init

How it works

knktx init detects which agents are installed on your machine
Configures the MCP server connection for each agent (HTTP client → knktx API)
Injects your rules and workflows into each agent's instruction file (CLAUDE.md, GEMINI.md, etc.)
Agents call knktx tools at session start to load project context and update state as they work

CLI Reference

The knktx CLI syncs your rules, workflows, skills, plugins, and MCP configs into local agent configuration files. No global install needed — run it with uvx.

> uvx knktx init

Bootstrap knktx in your project. Detects installed agents, configures MCP server connections, and injects rules and workflows into each agent's instruction file.

> uvx knktx update

Sync the latest rules, workflows, skills, plugins, and MCP configs from the knktx API into your local agent configuration files.

Ready to get started?

Open App Back to Home