Structured multi-agent workflow for OpenCode. Cheap models route, mid models build, premium models review. File locks, approval gates, persistent design artifacts, and mandatory verification on every task.
Requires OpenCode and Node.js 18+.
npx @jimzandueta/oowl installThe installer defaults to the low profile and walks you through model profile selection, then writes the framework files into your project.
Open OpenCode in the project and tell dispatcher what to build:
You: "Add a login page with email/password and Google OAuth."
Dispatcher → Architect writes design.md [you approve]
→ Planner writes implementation.md → plan-reviewer
→ [you approve the plan]
→ Builder schedules → agents implement with file locks
→ Reviewer writes review.md
→ [optional merge → done]
23 agents across 6 classes: orchestrators, artifact owners, implementers, reviewers, escalation, and bounded workers. Every substantial feature produces a docs/specs/<feature>/ directory with the design, plan, and review owned by their respective agents.
One agent designs, codes, and reviews in a single session. It edits files it shouldn't, makes changes you didn't approve, loses context when you scroll away, and never runs a security check unless you specifically ask. You get output fast, but you don't know what's safe without reading every diff.
oowl decomposes the work:
| Without oowl | With oowl |
|---|---|
| One agent touches anything | Each task has locked file boundaries |
| Design vanishes when chat ends | Design, plan, review persist in docs/specs/ |
| "Looks done" is the only check | Agents run verification before reporting complete |
| Same model does everything | Profiles assign cheap / mid / premium by role |
| Security if you remember | Sensitive areas auto-escalate |
| Review if you ask | Every feature ships with a written review |
your request → dispatcher
trivial? → one implementer → done
substantial?
→ optional Git feature branch
→ architect writes design.md [you approve]
→ planner writes implementation.md [plan-reviewer validates]
→ [you approve]
→ builder schedules implementation waves
→ agents implement with file locks and verification
→ reviewer writes review.md
→ optional branch merge → done
Artifacts are the durable record of every feature. They live under docs/specs/<feature>/:
docs/specs/login-page/
├── design.md # architect: approach, tradeoffs, risks
├── ui-spec.md # designer: layout, interactions (if UI involved)
├── implementation.md # planner: task breakdown, file locks, verification
└── review.md # reviewer: what shipped, test results, risks
These are owned by specific agents. Non-owners can read them but cannot edit, delete, or overwrite them.
Tiers map agents to three model cost levels:
| Tier | Handles | Agents |
|---|---|---|
| Cheap | Routing, scheduling, bounded work | dispatcher, builder, low-* |
| Mid | Design, implementation, review | architect, planner, specialists |
| Premium | Escalations, deep security | high-*, security-auditor |
Switch between bundled profiles (low default, balanced, high, free) or build a custom one from your connected models with oowl profile. The free profile uses free OpenCode models; free-data will be used in training, so do not use it for private, proprietary, regulated, customer, or confidential data.
| Command | What it does |
|---|---|
oowl install |
Install the framework (interactive walkthrough) |
oowl init |
Configure Optional skills per project |
oowl profile |
Switch model profiles interactively |
oowl profile <name> |
Switch directly: free, low, balanced, high, custom |
oowl update |
Update framework files with conflict detection |
oowl --version |
Print version |
Workflow, agent reference, install options, model profiles, customization, common issues, and contributing guide → https://jimzandueta.github.io/oowl/docs/.
MIT. See LICENSE.