Summary

OpenASE is an all-in-one platform for ticket-driven automated software engineering, transforming tickets into working code via AI agents. It targets engineers and power users seeking automated development pipelines with full traceability, offering a unified system for managing complex tasks with integrated human oversight.

How It Works

This project is a single Go binary combining an API server, orchestrator, and embedded web UI, eliminating runtime Node.js dependencies. Its core operates on a ticket-driven model: AI agents claim tickets based on status triggers and execute defined workflows on designated machines, logging all activity for traceability. It supports asynchronous "Ticket Agents" for autonomous execution and synchronous "Project AI" for exploratory conversations. Reusable "Skills" extend agent capabilities, managed via an integrated editor.

Quick Start & Requirements

Installation is via a one-command script: bash curl -fsSL https://raw.githubusercontent.com/pacificstudio/openase/main/scripts/install.sh | sh. Building from source requires cloning, make build-web, and ./bin/openase setup. Key prerequisites include Go and AI agent CLIs (Claude Code, Codex recommended). PostgreSQL is optional. The frontend build needs ~8 GB RAM. Control plane at http://127.0.0.1:19836. User guides available in English/Chinese.

Highlighted Details

• Ticket Orchestration: Kanban, list views, dependencies, custom statuses, repo scoping.
• Multi-Agent: Integrates Claude Code, Codex, Gemini CLI.
• Workflow Engine: Markdown "Harness" docs, skill binding, scheduled cron jobs.
• Security: OIDC login, RBAC, agent token auth.

Maintenance & Community

Roadmap includes Remote Runtime Operations, Windows Support. No specific community channels (Discord, Slack) or notable contributors/sponsorships mentioned.

Licensing & Compatibility

License type detailed in LICENSE file, not explicitly in README. Intended for local dev, private networks, trusted environments; not for public deployment.

Limitations & Caveats

Agents run with permissive flags by default, granting broad host command execution; use only on trusted infrastructure. Native Windows support and cross-org collaboration are work-in-progress.