OpenCode Onboard: One-Click AI Agent Workflow Scaffolding Tool, Turn Codebases into Agent-Friendly Projects in Seconds

OpenCode Onboard is a command-line tool that configures a complete AI agent workflow for any codebase with a single command. It can automatically install a general-purpose agent team, configure multi-model strategies, initialize OpenSpec change management, and integrate the Ensemble parallel execution framework. It solves the problem of codebases being unfriendly to AI agents and provides a standardized solution for team collaboration and automated development.

Traditional codebases lack AI-designed documentation (e.g., AGENTS.md), clear task assignment mechanisms, and agent-readable context information. This causes AI tools to only complete simple code completion tasks and struggle to undertake complex development tasks. The core idea of OpenCode Onboard is to proactively transform codebases into an agent-friendly form, just like preparing onboarding materials for new developers.

Complete the 10-step configuration via an interactive command-line wizard: 1. Define source code scope; 2. Clean up old AI configurations; 3. Select code hosting platform (GitHub/Azure DevOps); 4. Verify platform CLI; 5. Copy scaffolding files (core step); 6. Initialize OpenSpec; 7. Select AI models (three categories: planning/building/fast); 8. Token optimization tool; 9. Install browser plugin; 10. Write configuration metadata.

Adopts an architecture that separates agents (defining "how to work") from skills (defining "what to know"). The default agent team includes devops-manager (coordinator) and basic-engineer (executor); the skill system automatically discovers SKILL.md under .agents/skills/, with built-in skills like ob-global and ob-default, and supports custom skills.

Models are divided by role: planning models for complex reasoning, building models for code implementation, and fast models for auxiliary tasks. All models are labeled with cost levels ([$] economical, [$$] standard/high-end), providing cost transparency to help teams make economically rational choices.

After inputting a work item URL, the process is: load ob-global rules → parse work item → generate proposals/specifications/tasks → parallel execution (based on the Ensemble framework, each agent in an independent git worktree) → verification → PR creation. Task status can be monitored in real-time via localhost:4747.

Provides custom slash commands: /init (initialize project), /plan (generate task list), /main (quick implementation). Supports incremental updates: clean (reset old configurations), copy (update templates/skills), optimization (configuration optimization), metadata (refresh snapshot).

Suitable for teams wanting to systematically introduce AI agents, multi-project organizations, cost-sensitive developers, and enterprises with DevOps integration needs. In the future, it will become an important part of software development infrastructure, helping teams efficiently utilize AI capabilities while maintaining process controllability.