@openclaw/orchestration is a multi-agent task coordination framework by frank-bot07 that uses a shared SQLite database as a task queue. It provides atomic task claiming, dependency management, timeout handling, and markdown-based task visibility, making it possible to coordinate multiple AI agents working on related tasks without conflicts or duplication.

Key Features

SQLite-backed task queue -- A single SQLite database serves as the source of truth for all task state, eliminating the need for external message brokers or queue services.
Atomic task claiming -- Ensures only one agent can claim a given task at a time, preventing duplicate work across concurrent agents.
Task dependency management -- Define sequential workflows where tasks must complete in order, with downstream tasks automatically unblocked when dependencies are satisfied.
Timeout and retry handling -- A sweep mechanism detects stalled tasks that have exceeded their timeout and makes them available for retry by other agents.
Markdown interchange files -- Task state is mirrored to markdown files that provide human-readable visibility into queue status, making it easy to monitor progress across agents.
CLI interface -- A Node.js command-line tool supports registering agents, creating tasks, claiming assignments, marking completion, and refreshing the markdown interface.

Use Cases

Multi-agent code generation -- Coordinate multiple AI agents working on different parts of a codebase, with dependency tracking ensuring that foundational modules are built before dependent features.
Pipeline orchestration -- Build multi-step processing pipelines where each stage is handled by a specialized agent, with automatic handoff between stages.
Parallel task distribution -- Distribute independent tasks across multiple agents with atomic claiming to maximize throughput while preventing duplicate effort.
Supervised automation -- Use the markdown interchange files as a monitoring dashboard to review agent progress and intervene when tasks fail or stall.

How It Works

The framework maintains a SQLite database where tasks are registered with metadata including dependencies, timeouts, and agent assignments. Agents claim tasks atomically, execute them, and report completion. A periodic sweep process identifies timed-out tasks and returns them to the available pool. All state changes are reflected in both the database and companion markdown files for visibility.

Getting Started

Navigate to the skill directory and run
```
npm install
```
to install dependencies.

node cli.js register-agent --name "agent-1"

Create a task:

node cli.js create-task --title "Build auth module" --timeout 300

Claim and execute:

node cli.js claim --agent "agent-1"

followed by

node cli.js complete --task-id <id>

Monitor status via the generated markdown files in the workspace directory.

Note: This skill has been flagged for potential race conditions in database transactions. It is recommended for development and prototyping rather than production use without additional hardening.

@openclaw/orchestration

AI Skill Market Insights

Be Part of the 0+ Developer Community

Key Features

Use Cases

How It Works

Getting Started

License

Quick Start

Manual Installation

TEAR & SHARE

Tags

HTML to PowerPoint Guide

PDF Processing Advanced Reference

Office Open XML Technical Reference

Fillable fields

DOCX Library Tutorial

Channels

Learn

Compare

Company