🧰 opencode-onboard

Prepare any codebase for AI. Wires OpenCode, OpenSpec, codegraph, and basic-memory into a multi-agent development workflow powered by native parallel subagents.

GitHub, Azure DevOps, Jira, GitLab, browser-based backlog, or combinations (e.g. Jira backlog + GitHub repo, Others (Browser) + GitLab repo).

opencode.ckgrafico.com

What is this?

Most codebases have no AGENTS.md, no architecture docs agents can read, and no defined workflow for picking up tasks. Agents end up improvising, producing inconsistent results.

opencode-onboard fixes that in a single interactive wizard. It configures OpenCode with OpenSpec for structured change management, native subagent waves for parallel agent execution, codegraph for code intelligence, and basic-memory for shared context across agent sessions. It also installs an agent team, platform skills, and slash commands — everything agents need to plan, implement, and ship.

Quick start

npx opencode-onboard@latest

Requires Node.js 18+.

Run specific steps

You can run individual setup/maintenance steps without running the full wizard:

# Run one step directly
npx opencode-onboard clean
npx opencode-onboard platform
npx opencode-onboard copy
npx opencode-onboard openspec
npx opencode-onboard models
npx opencode-onboard optimization
npx opencode-onboard browser
npx opencode-onboard metadata
npx opencode-onboard join

# Show CLI help and all commands
npx opencode-onboard --help
npx opencode-onboard -h

When available, step commands reuse context from .opencode/opencode-onboard.json.

Typical flow for reruns:

Run clean if you want to reset old AI files
Run copy if templates/skills changed in a new onboard release
Run optimization if you want to reconfigure RTK/quota/caveman and the AGENTS.md optimization markers
Run metadata last to refresh .opencode/opencode-onboard.json
Run join if you're a new member of an existing onboarded project and want to sync the latest onboarding metadata

How it works

The CLI runs a 10-step onboarding wizard. It keeps the current step visible, plus the last two completed steps, so progress is always clear.

Step	What happens
1. Source scope	Choose current repo or sibling source roots for code analysis
2. Clean AI files	Detects existing `AGENTS.md`, `.cursorrules`, `CLAUDE.md`, `.agents/` etc. and removes them, preserves your `.agents/skills/`
3. Choose platform	Backlog (GitHub / Azure DevOps / Jira / Others (Browser) / None) + repo (GitHub / Azure DevOps / GitLab / None). Supports mixed platforms e.g. browser backlog + GitHub repo, Jira backlog + GitLab repo
4. Check platform CLI	Verifies `gh` (GitHub) or `az` + `azure-devops` (Azure DevOps) or `acli` (Jira) or `glab` (GitLab). Skips CLI checks for Others (Browser) or None
5. Copy scaffolding	Copies agents + built-in skills + bootstrap docs, writes source-roots metadata, applies AGENTS bootstrap patching, copies `skills-lock.json`, then runs `npx skills`
6. Init OpenSpec	Runs `npx @fission-ai/openspec init` silently for structured change management
7. Choose models	Fetches live model list from models.dev, lets you pick plan / build / fast models with cost indicators and canonical pricing
8. Token optimization tools	Optional (recommended). One checklist step for RTK check, opencode-quota setup, caveman install, and token-optimization rule injection into AGENTS.md + command files
9. Install browser plugin	Installs `@different-ai/opencode-browser` globally for agent browser automation
10. Write onboarding metadata	Writes `.opencode/opencode-onboard.json` with selected setup details

When it finishes, open OpenCode in your project and type:

/ob-init

OpenCode asks if this is a greenfield or brownfield project. For brownfield projects it generates ARCHITECTURE.md and DESIGN.md from your actual codebase, archives project history, then activates the full agent team. For greenfield projects it skips doc generation and leaves placeholder files you can populate later with /ob-create-architecture and /ob-create-design.

Commands

Custom slash commands are installed into .opencode/commands/ and are available directly in OpenCode.

Command	Description
`/ob-help`	Show all commands and when to use each one. Start here if you're unsure.
`/ob-init`	Initialize the project. Asks greenfield vs brownfield, then activates the agent team.
`/ob-explore`	Think through an idea or investigate a problem before committing to a plan.
`/ob-propose <url or idea>`	Parse a GitHub Issue / Azure DevOps / Jira / browser URL or a direct idea into a structured plan (proposal, specs, tasks). Enriches each task with agent and model assignments.
`/ob-apply`	Implement tasks from the current OpenSpec change via parallel subagent waves (native `task` tool).
`/ob-pullrequest`	Create a PR for the current branch, or read and classify PR review comments.
`/ob-archive`	Archive a completed OpenSpec change.
`/ob-main <task>`	Quick direct implementation — no OpenSpec, no waves, no PR. Just do it.
`/ob-autopilot <feature or URL>`	Autonomous, no-confirmation pipeline: branch off `main`, then propose → apply → archive (one commit per phase). Default: merge to `main` + delete branch. Add `push` keyword to push branch only. Add `pr` keyword to push + create a PR. For loop-engineering.
`/ob-create-engineer <name> <tier> "<description>"`	Create a custom specialist engineer with skills auto-installed from skills.sh. Agent file is a template; tier variants are injected by the `ob-subagent-tiers` plugin at startup.
`/ob-create-architecture`	Generate or regenerate `ARCHITECTURE.md` from the codebase.
`/ob-create-design`	Generate or regenerate `DESIGN.md` from the design system.
`/ob-create-project-guardrails`	Generate a `project-guardrails` skill from `ARCHITECTURE.md` + project config files. Extracts architecture boundaries, naming, code style, testing, and git workflow rules. Updates all `*-engineer.md` to load the skill.
`/ob-set-model [user] <tier> <model>`	Set the model for a tier (`plan`, `build`, `fast`). Writes to `opencode-onboard.json` (team) or `opencode-onboard.user.json` (user override, gitignored) when `user` prefix is used. Restart to pick up — the `ob-subagent-tiers` plugin rebuilds tier agents at startup. Pass a model id or `current` for the active session model.

Agents and Skills

opencode-onboard draws a hard line between two concepts:

Agents, universal behaviors

Agents define how to work. They are universal personas (same behavior across projects and stacks).

Current baseline uses a generic execution model:

lead     lead/orchestrator, planning, PR lifecycle
basic-engineer     implementation worker, ability-driven

basic-engineer behavior is composed by abilities, not hardcoded role silos. Project-specific specialization comes from user-created custom engineers via /ob-create-engineer. During /ob-apply, the lead should inspect the engineers that actually exist in .opencode/agents/, prefer matching custom engineers, and fall back to basic-engineer only when no specialist is a clear fit.

Skills, platform knowledge

Skills define what to know. They provide project rules, platform behavior, and task-specific execution guidance. Agents auto-detect/load relevant skills; you do not manually choose skills per prompt.

If you choose backlog platform None, no userstory skills are injected into the workflow. The project works from direct conversation, local repo context, and optional OpenSpec artifacts only. If you choose repo platform None, no pull-request skills are injected.

Current loading model:

ob-generic-guardrails is mandatory baseline for every agent (git/secrets/quality rules + the engineer workflow)
ob-default is fallback when nothing else matches
Baseline context rules and token-optimization guidance live in AGENTS.md (always in context), not in a skill

Default basic-engineer abilities:

## Abilities
- Guardrails: @ob-generic-guardrails, @ob-default
- Development: @ob-default
- Testing: @ob-default
- Infrastructure: @ob-default

Users are expected to create additional skills and map them into abilities over time.

Built-in skills (ob- prefix) shipped with opencode-onboard:

Skill	Purpose
`ob-default`	Fallback, when no other skill matches
`ob-generic-guardrails`	Foundation for user guardrails skills
`ob-userstory-gh`	Parse a GitHub Issue URL into a structured work item
`ob-userstory-az`	Parse an Azure DevOps work item URL
`ob-userstory-jira`	Parse a Jira issue URL via `acli` CLI
`ob-userstory-browser`	Parse work item from any URL via browser automation (Linear, Trello, etc.)
`ob-pullrequest-gh`	Create GitHub PRs with screenshots, triage review feedback
`ob-pullrequest-az`	Create Azure DevOps PRs, triage review feedback
`ob-pullrequest-gl`	Create GitLab merge requests, triage review feedback
`browser-automation`	Browser control via `@different-ai/opencode-browser` (localhost + browser backlog exception)

Skills live in .agents/skills/. Any SKILL.md file in a subdirectory is automatically discoverable, write your own and agents will pick them up.

Models, plan / build / fast

During onboarding you pick three models:

Role	Used by	Pick
plan	Main OpenCode session (the lead)	Something capable with strong reasoning
build	Specialist engineers (default tier)	Something capable for implementation
fast	`basic-engineer` & light helpers	Something fast and cheap

Models are fetched live from models.dev (3000+ models, cached weekly). Cost tiers [$] [$$] [$$$] always reflect the canonical provider price, so github-copilot/claude-opus-4.7 shows [$$] not [$].

The pipeline

When you give the lead agent a work item URL, execution follows this pipeline. If backlog platform is None, skip the work item stage. If repo platform is None, skip the PR stage:

lead
                  ↓
         parse work item via userstory skill
                  ↓
              openspec-propose
        proposal + specs + tasks
                  ↓
             [confirm with user]
                  ↓
 wave of subagents (basic-engineer / *-engineer, per-tier model)
 each implements its assigned tasks → returns result → lead commits group
                  ↓
       verify (tests/build/lint as needed)
                  ↓
    lead (ship mode, if configured)
  commit → push → PR → feedback loop

Load the platform userstory skill (installed as ob-userstory, from the variant matching your backlog platform)
Run /ob-propose to produce proposal.md, specs, and tasks.md
Confirm with user before implementation
Run /ob-apply to orchestrate implementation in waves
Each wave spawns engineers in parallel (basic-engineer and/or custom engineers, each carrying its own tier model), capped at maxConcurrentAgents
Each subagent receives its task IDs in its prompt, loads relevant abilities, implements, and returns; the lead commits each group
Verify with tests/build/lint according to task scope
Ship/update PR via lead flow

Agents run as native OpenCode subagents in parallel waves — no external plugin, no git worktrees. The lead's Todo pane is the live board, and the ob-subagent-monitor plugin mirrors state to .opencode/.ob-run.json. Navigate into any running subagent with ctrl+x ↓ then ←/→.

What gets installed

your-project/
├── AGENTS.md                        ← bootstrap mode, replaced after first "/ob-init"
├── ARCHITECTURE.md                  ← prompt for agents to fill in from your codebase
├── DESIGN.md                        ← prompt for agents to fill in from your codebase
├── .opencode/
│   ├── opencode.json                ← default model + plugin config
│   ├── opencode-onboard.json        ← onboarding metadata + runtime config (models, maxConcurrentAgents)
│   ├── agents/                      ← basic-engineer + user-created *-engineer files (each carries its model)
│   ├── tui.json                     ← registers the Subagents sidebar panel
│   ├── tui/
│   │   └── ob-subagents.tsx         ← TUI plugin: live Subagents panel in the sidebar
│   └── plugins/
│       └── ob-subagent-monitor.js   ← server plugin: writes subagent state → .opencode/.ob-run.json
└── .agents/
    └── skills/
        ├── ob-default/             ← fallback skill
        ├── ob-generic-guardrails/  ← foundation for user guardrails
        ├── ob-userstory/           ← the variant matching your backlog platform, renamed on install
        ├── ob-pullrequest/         ← the variant matching your repo platform, renamed on install
        └── browser-automation/

Platform skills ship as suffixed variants (ob-userstory-gh/-az/-jira/-browser, ob-pullrequest-gh/-az/-gl) and the installer copies only the matching one, renamed to its generic name. Source-roots metadata lands in .opencode/source-roots.json; token-optimization guidance is injected into AGENTS.md marker blocks during onboarding.

The bootstrap sequence

The first time you type init in OpenCode after onboarding, the agent asks whether this is a greenfield or brownfield project:

Brownfield (existing codebase)

Bootstrap-mode AGENTS.md triggers the initialization workflow
OpenCode archives existing project context into OpenSpec (project-history)
OpenCode runs /ob-create-architecture → generates real ARCHITECTURE.md from your codebase
OpenCode runs /ob-create-design → generates real DESIGN.md from your design system
OpenSpec config.yaml is populated with discovered tech stack and domain context
Bootstrap AGENTS.md is replaced with production guidance
Team workflows become fully active for normal implementation tasks

Greenfield (new project, little or no existing code)

Bootstrap-mode AGENTS.md triggers the initialization workflow
OpenSpec config.yaml is populated with what is known (intended stack, domain)
Bootstrap AGENTS.md is replaced with production guidance
ARCHITECTURE.md and DESIGN.md are left as placeholder files

Once your codebase has meaningful content, run:

/ob-create-architecture to generate architecture docs
/ob-create-design to generate design system docs

Both commands are safe to rerun at any time as the project evolves.

Token Budget Controls

Long unattended agent sessions can consume significant tokens. Set these controls up before first use:

Set provider-side limits first — monthly soft-limit + hard usage cap in your provider dashboard:
- OpenAI: platform.openai.com/account/limits
- Anthropic: console.anthropic.com
- Google AI Studio: aistudio.google.com/app/usage
Route models by task type — use a fast/cheap model (e.g. haiku, gpt-4o-mini) for orchestration and status loops; reserve expensive models (e.g. sonnet, opus, gpt-4o) for implementation tasks only.
Install the quota plugin — the @slkiser/opencode-quota plugin adds /quota and /quota_status commands that surface real-time token usage inside OpenCode sessions.
Use /quota checkpoints — run /quota before starting any /ob-apply session and after each agent wave. Pause at 75% consumed; stop at 90%.
Confirm before large runs — the onboarded /ob-apply workflow will ask for your confirmation before spawning agents for Medium (4–7 tasks) or High (8+ tasks) scope sessions.

Prerequisites

Requirement	Notes
Node.js 18+	Required
OpenCode	The agent runtime
gh CLI	GitHub platform, must be authenticated
az CLI + azure-devops extension	Azure DevOps platform
acli	Jira (Atlassian) backlog platform, must be authenticated
glab	GitLab repo platform, must be authenticated

Development

Wizard choices and defaults live in src/presets/ where possible:

source.json controls source-scope prompt options
platforms.json controls platform labels, CLI checks, and backlog-only flags
clean.json controls AI file detection and preservation
models.json controls model role prompts and agent assignments
optimization.json controls RTK/quota/caveman checklist defaults
quota.json controls opencode-quota defaults
browser.json controls opencode-browser installer automation

git clone https://github.com/ckgrafico/opencode-onboard.git
cd opencode-onboard
pnpm install

# Run the CLI locally
node src/index.js

# Run tests
pnpm test

# Run linting
pnpm lint

# Fix auto-fixable lint issues
pnpm lint:fix

# Watch mode
pnpm test:watch

Tests are written with Vitest. Linting uses ESLint flat config with Node ESM defaults and stricter correctness rules.

Name		Name	Last commit message	Last commit date
Latest commit History 181 Commits
.github		.github
content		content
docs		docs
src		src
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
demo.gif		demo.gif
eslint.config.js		eslint.config.js
jsconfig.json		jsconfig.json
logo.png		logo.png
package.json		package.json
pnpm-lock.yaml		pnpm-lock.yaml

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Repository files navigation

🧰 opencode-onboard

What is this?

Quick start

Run specific steps

How it works

Commands

Agents and Skills

Agents, universal behaviors

Skills, platform knowledge

Models, plan / build / fast

The pipeline

What gets installed

The bootstrap sequence

Brownfield (existing codebase)

Greenfield (new project, little or no existing code)

Token Budget Controls

Prerequisites

Development

License

About

Uh oh!

Releases 7

Sponsor this project

Uh oh!

Contributors

Uh oh!

Languages

Uh oh!

Folders and files

Latest commit

History

Repository files navigation

🧰 opencode-onboard

What is this?

Quick start

Run specific steps

How it works

Commands

Agents and Skills

Agents, universal behaviors

Skills, platform knowledge

Models, plan / build / fast

The pipeline

What gets installed

The bootstrap sequence

Brownfield (existing codebase)

Greenfield (new project, little or no existing code)

Token Budget Controls

Prerequisites

Development

License

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 7

Sponsor this project

Uh oh!

Contributors

Uh oh!

Languages