SAGE AI
First learn, then vibe~
Sage. A new tool to guide us all.
Automation abstracts the knowledge you need. Sage is the counter-spell — a pocket genie in your terminal that reads your project, explains the why behind the code, and keeps you present while you build.
Learn first, then VibeCode
Don't let AI deprive you of learning. Stay in your editor, invoke the genie, and keep growing while you build.
Explore new frontiers
Learn new stacks, sharpen your practices, and let /wish remember every step of your journey across sessions.
Read-only by design
Sage doesn't touch your code. It maps it, traces it, explains it. Your repo stays exactly as you left it.
/exp-file @apps\backend\src\users\users.controller.ts The Users controller exposes 5 REST endpoints but only 2 are implemented (create and getAll). The other 3 throw NotImplementedException from the service. Key points Role HTTP layer of the Users module, delegates to UsersService Real endpoints POST / (create), GET / (getAll) Stubs GET /:_id, DELETE /:_id, PATCH /:_id Security No @UseGuards despite Auth/Users tag Convention _id prefix parameter can be misleading What the agent found - CRUD ahead of implementation — interface exists, backend incomplete. - No endpoint protection (no guards, no authentication). - Design phase clue in ADR002. Explore the service or check ADR002 directly with /why?
How it works
- [ 1 ]Install Sage✨ [ Global or Local ]
- [ 2 ]Launch OpenCode
- [ 3 ]Select Sage as sub-agent — press Tab in OpenCode
- [ 4 ]Simply run /exp, /exp-file, /flow or /why and start exploring your project.
- [ 5 ]Want a read on safety or craft? Run /sec for heuristics on threats, or /audit for a mentor-style craft audit — they hand off to each other.
- [ 6 ]Ready to learn? Start with /wish and begin your study or implementation plan — without ever leaving OpenCode!
Available /Commands
- /exp
Scans the entire project and produces a structured overview: stack, architecture, entry points, and key conventions. Also creates/updates sources.json so other commands have context.
- /exp-file@path
Explains a single file in depth — what it does, why it exists, and how it connects to the rest of the project.
- /flow@path
Maps the dependency graph of a module — what it imports, what imports it, and how data flows through it.
- /why@path
Traces the design decisions behind a file using git history, ADRs, and inline patterns. Teaches you to read intent, not just code.
- /wish
Your co-op mentor for conquering new knowledge. Tackling a stack you've never touched? Learning a tech that feels out of reach? Set a goal, pick a level (Casual, Formal, Senior), and Sage walks you through it step by step across sessions — remembering your progress, your decisions, and where you left off.
- /sec@path
Sage reviews your file or project with a focus on security and flags up to 5 things worth a second look — exposed secrets, doors left open, risky config. Each finding is rated 🔴 High · 🟡 Medium · 🟢 Low, always as a concern, never a scolding. A quick gut-check to catch the obvious — not a replacement for a real security audit.
- /audit@path
Sage audits and evaluates your file or project for quality and best practices — first naming what's genuinely well done (and why), then suggesting small, optional improvements in style, maintainability and modern patterns. Suggestions read as "if you'd like, you could…", never orders. /sec and /audit are complementary views: one watches for threats, the other for craft — each hands off to the other when it makes sense.
*Responses may vary depending on the model in use.
**Run /exp with a strong planning model — it's the heaviest command, optimized to avoid back-to-back re-runs.
***@path refers to a specific file — e.g. index.js, app.tsx, or any path in your project.
Built for more.
SAGE AI integrates with the skills available in your OpenCode project — and works with any model, from free tiers to paid plans, all the way to your own self-hosted LLM.