AutoClaw: Persistent Multi-Agent Development Assistant in VS Code

AutoClaw is a VS Code extension that introduces the concept of persistent background agents to address the pain points of the "on-demand response" mode of existing AI programming assistants. Its core capabilities include KDream background memory agent, AutoBuild automated workflow, and MAteam multi-agent coordinator. It supports mainstream AI tools like Copilot, Claude Code, and Cursor, enabling zero-configuration plug-and-play.

Existing AI programming assistants like GitHub Copilot and Claude Code use an "on-demand response" mode, only intervening when developers actively initiate a conversation. This makes it difficult to meet the continuous tracking needs of complex projects: TODO items in the codebase are easily forgotten, there's a lack of reminders for long-uncommitted code, and cross-session context is hard to maintain. AutoClaw integrates AI assistants into daily development workflows through persistent background agents to solve these pain points.

AutoClaw is built around three core capabilities:

KDream: Always-On Background Memory Agent

Adopts a Tick-driven model to continuously monitor Git status, TODO markers, and to-do items. It maintains .autoclaw/kdream/memory/MEMORY.md to record project knowledge and periodically performs "Dream" memory integration. It supports three task injection methods: code marking, command addition, and manual editing.

AutoBuild: Self-Scheduled Build Workflow

Implements in-process scheduled task scheduling based on node-cron, supporting scenarios like local testing, document generation, dependency checking, and data synchronization.

MAteam: Multi-Agent Coordinator

Implements a collaborative pipeline of Researcher→Coder→Reviewer→Verifier. It launches sub-agents for parallel processing in Claude Code, while using in-session simulation for other tools.

AutoClaw supports mainstream AI programming assistants. Upon first activation, it automatically detects installed extensions and copies the corresponding skill files:

AI Tool	Skill Loading Method
GitHub Copilot	VS Code built-in chatSkills
Claude Code	Copy skill files to ~/.claude/skills/
Cursor	Copy .mdc rules to .cursor/rules/
Kiro	Copy steering files to .kiro/steering/
Windsurf	Copy rules to .windsurf/rules/
KiloCode	Merge custom modes into .kilocodemodes
Cline	Copy rules to .clinerules/
Continue	Copy .prompt files to .continue/prompts/

AutoClaw provides extension-level commands:

• AutoClaw: Doctor: One-click generation of a health report covering workspace status, KDream running status, MEMORY.md integrity, etc.
• AutoClaw: Export Health Snapshot: Export a snapshot in Markdown format for easy sharing and troubleshooting.
• AutoClaw: Run AutoBuild Workflow Now / Tail Latest Workflow Run: Manually trigger the workflow and view real-time logs.

The KDream Dashboard provides an intuitive interface showing:

• Status Panel: KDream running status, Tick count, last Dream time
• Tasks and To-Dos: Tasks in MEMORY.md and their completion status
• Recent Activities: Last 10 lines of the day's logs
• Adapter Health: Integration status of each AI extension
• TODO List: TODO items marked in the code

It can be opened via the lobster icon in the activity bar or the command palette.

Start KDream: Enter /kdream start in Claude Code/Copilot/Continue; for tools like Cursor, directly describe your needs to activate the rules.

Daily Commands:

• /kdream ps: Check running status
• /kdream logs: View the day's logs
• /kdream stop: Close the agent
• /kdream dream: Integrate memory immediately
• /kdream work <task>: Delegate a task

Workflow Recommendations:

1. Project Initialization: Start KDream to scan existing TODOs
2. Daily Development: Add TODO comments for KDream to track
3. Task Handover: Use /kdream work to delegate tasks
4. Regular Review: Check the Observations section in MEMORY.md to understand project evolution.

AutoClaw transforms AI from a "passive responder" to an "active collaborator", making it suitable for teams looking to improve development efficiency and code quality management. Licensing Model:

• Personal, educational, and evaluation use: Free
• Commercial use: License purchase required (contact ZippyTechnologiesLLC)

Outlook: Represents an important direction for the evolution of AI-assisted development towards "active intelligence".