DOCSEARCHbyAlgolia
Submit Docs

cckit — a project OS for coding agents

website·Indexed 30 days ago

https://cckit.vercel.app

A CLI tool and project OS designed to drive the full GitHub work lifecycle for coding agents, specifically optimized for Claude Code. cckit provides a structured framework for managing efforts, waves, and worktrees, featuring a security secret guard, persistent memory via MemPalace, and a system of hooks and skills to automate the path from GitHub issues to merged PRs.

Tokens
33.6K
Snippets
176
Records
282
Agent score
67%

What's inside cckit

  1. Overview of cckit's purpose and core philosophy

    cckit is designed to turn a Git repository into retrievable, efficient context by automating the structural mechanics of software development. It handles the 'structural engineering'—including the management of branches, worktrees, pull requests, and issue trails—allowing the developer to act as the 'architect' and idea generator. The tool integrates issues, GitHub Projects, and advanced Git mechanics to maintain a disciplined, SCRUM-style tracking process without the manual overhead.

  2. Overview of cckit and its purpose

    cckit is a tool designed to turn a Git repository into retrievable, efficient context for development. It acts as a 'structural engineer' by automating the mechanics of the development process—such as managing branches, worktrees, pull requests, and issue tracking—allowing the developer to focus on being the architect and idea generator. It integrates issues, GitHub Projects, and advanced Git mechanics to maintain a disciplined, SCRUM-style development lifecycle.

  3. Overview of cckit for coding agents

    cckit is a project OS and CLI tool designed to manage the full GitHub work lifecycle for coding agents, such as Claude Code. It transforms a Git repository into retrievable, efficient context for agents, allowing them to handle the mechanics of development—including issue tracking, worktree management, and PR creation—while the developer provides the ideas and architecture.

  4. Overview of MemPalace persistent memory in cckit

    MemPalace is an optional, local, no-API-key engine that provides persistent memory for coding agents. It mines projects and conversations into searchable "wings" (scoped per project), preventing agents from being stateless. It reduces token consumption by providing a wake-up context of ~600–900 tokens, ensuring continuity of decisions, ADRs, and project architecture across different sessions.

  5. Overview of cckit project OS

    cckit is a project OS designed for coding agents (such as Claude Code). It provides a CLI that manages the full GitHub work lifecycle, transforming Git history and issue tracking into retrievable, efficient context for agentic development. It allows developers to automate the mechanics of shipping work—from managing worktrees and PRs to running parallel issue waves—while maintaining security guards and persistent project memory.

  6. View the cckit status dashboard

    The status command provides a thin dashboard overview of the board, worktrees, and handoff state.
    cckit status

  7. Understand cckit documentation tags

    The cckit documentation uses a visual tagging system to indicate how a feature is triggered or its nature. These tags appear next to headings or names throughout the guide to help users quickly identify actionable items without reading the full text.

  8. Understand the cckit architecture split

    cckit is divided into two distinct layers to ensure portability and agent-agnosticism:

    1. Core: A pure bash implementation that owns all operations (GitHub lifecycle, worktrees, orchestration, etc.). It depends on no specific agent and uses a standard toolchain (git, gh, jq, tmux).
    2. Adapter: A thin layer specific to an agent (e.g., Claude Code) that provides discovery and ergonomics to call the core operations.

    This split ensures that operations are implemented once in the core, preventing logic drift and allowing cckit to run in CI, cron jobs, or bare terminals without an AI agent.

  9. Debug web pages using cckit debug

    The cckit debug command provides AI agents with a scriptable, token-efficient method to debug web pages. It captures screenshots, accessibility snapshots, console logs, network logs, and Lighthouse reports. It achieves this by driving chrome-devtools-axi, which emits output in TOON (a compact text format) to prevent bloating the agent's context window.
    cckit debug <axi-args...>

  10. View the cckit status dashboard

    The status command provides a thin dashboard overview of the board, worktrees, and handoff state.
    cckit status

  11. Identify cckit trigger and concept tags

    Use the following tag definitions to determine how to interact with a feature in cckit:

    User-Triggered Actions

    • Conversational: Triggered by speaking/typing plain language to Claude.
    • Command: A CLI verb executed in a terminal (e.g., cckit ...).
    • Skill: A slash command (e.g., /skill-name) executed inside Claude Code.

    Automated Actions

    • Hook: Runs automatically in response to a specific event.

    Conceptual Information

    • Flow: A defined end-to-end process.
    • Agent: Machine-facing output (e.g., --llm or TOON).
    • Concept: A fundamental term or prerequisite knowledge.

  12. Understand the cckit architecture: Core vs Adapters

    cckit is architected into two distinct layers to ensure that operations remain agent-agnostic while providing native experiences for specific AI agents.

    • Core: A pure bash implementation that owns every operation (GitHub lifecycle, worktrees, orchestration, etc.). It has no dependency on any specific agent and can run in CI, cron jobs, or a bare terminal.
    • Adapter: A thin layer specific to an agent (e.g., Claude Code) that allows the agent to discover and call the core operations. Adapters do not re-implement logic; they only route the agent to the core verbs.