OpenASE: Ticket-Driven Automated Software Engineering Platform, Full Lifecycle Management of AI Agents

Introducing OpenASE, an integrated platform that converts tickets into working code, supporting multi-agent CLI, synchronous and asynchronous human-AI collaboration, complete traceability, and providing enterprise-grade infrastructure for AI-driven software development.

With the leap in AI coding assistant capabilities, software development is undergoing profound changes. However, truly integrating AI tools into team workflows is not easy: How to coordinate multiple AI agents? How to ensure the controllability of human-AI collaboration? How to track the decision-making process of AI? These issues require systematic solutions, not just better models.

OpenASE (Open Automated Software Engineering) is an ambitious open-source project aimed at building a ticket-driven automated software engineering platform. It deeply integrates traditional project management (tickets, kanban, workflows) with modern AI agent technology, providing a complete control plane to manage the entire lifecycle of AI-assisted development.

OpenASE's design philosophy can be summarized as "Ticket as Code"—every development task is a ticket, each with a clear workflow definition, and AI agents automatically trigger execution based on the ticket's status. This model brings several key advantages:

• Traceability: Every development decision is linked to a specific ticket, facilitating auditing and review
• Controllability: Control the behavioral boundaries of AI through workflow definitions and constraint documents (Harness)
• Collaborability: Supports human-AI collaboration and multi-person collaboration; AI agents can participate in development as team members
• Scalability: Modular architecture supports multiple AI CLI tools and custom skills

OpenASE adopts a single-binary architecture, packaging the API server, workflow orchestrator, and embedded Web UI into a single Go binary file. This design simplifies deployment and operation while maintaining functional integrity.

Core components include:

• Ticket System: Kanban view, list view, dependency tracking, parent-child ticket relationships
• Agent Platform: Supports multiple AI tools such as Claude Code, Codex, Gemini CLI
• Workflow Engine: Harness document editing, state binding, skill binding
• Skill System: Built-in and custom skills to extend agent capabilities
• Machine Management: Multiple machine types including local, direct connection, reverse connection
• Activity Stream: Real-time SSE event stream that fully records the execution process
• Scheduled Jobs: Cron-based ticket creation and triggering

One of OpenASE's unique features is that it clearly defines two modes of human-AI collaboration:

When requirements are clear, acceptance criteria are well-defined, and the Harness constraint document is configured, the Ticket Agent can independently execute the entire task. It follows workflow instructions, updates ticket status, and completes the work described in the ticket—without continuous human supervision.

The workflow defines state transitions and the agent's behavior at each stage. Common patterns include:

• Full-Stack Developer Mode: A single agent handles the complete lifecycle (Todo → In Progress → In Review → Merging → Done)
• Hybrid Relay Mode: Multiple specialized agents collaborate (Design → Backend → Frontend → Testing → In Review → Merging → Done)

When requirements are vague, technical solutions need to be explored, or a formal ticket is not yet ready, you can have a synchronous conversation with Project AI. Project AI is an interactive assistant in the control plane's sidebar, which can:

• Analyze requirements and explore technical solutions
• Draft PRDs and documents
• Initialize code repositories
• Directly modify workflows, Harness, and skills
• Trigger agent execution
• Control Git operations (commit, branch, push)

Each Project AI tab runs in an independent workspace, allowing multiple agents to run in parallel without interfering with each other.

Harness is OpenASE's core security mechanism. It is a Markdown-formatted constraint document that defines the behavioral boundaries of AI agents. Unlike system prompts, Harness is a hard constraint—agents must comply with its rules.

Harness can define:

• Allowed and prohibited operations
• Code style and architecture constraints
• Testing and documentation requirements
• Security review checklists

This design allows teams to enjoy the convenience of AI automation while maintaining control over code quality and security.