A mini Applicant Tracking System (ATS) — a proof of concept: can a squad of AI agents build a real software project end-to-end, just like a human team? And can a second squad handle everything non-code — research, content, analysis — in parallel, on a different AI runtime?
Every line of code and every doc in this repo was written by AI agents operating as structured squads inside Multica.
A kanban-style ATS with 7 hiring stages: Applied → Screening → Interview → Offering → Negotiation → Hired → Not Proceeding
Features:
- Position management (create, filter, delete with safety check)
- Candidate CRUD with full profile fields
- Drag & drop between stages
- Dashboard metrics (Screening Rate, Selection Rate, Acceptance Rate, Avg Time to Hire)
- AI chat assistant — reads live pipeline data, scoped to HR/recruitment topics
- Day / Night mode (persists via localStorage)
- Real-time search + position filter
- Next.js 16 (App Router) · React 19 · Tailwind v4 · TypeScript strict
- JSON file store — no external DB, zero config
npm install
npm run dev # http://localhost:3000
npm run build # type check + production build| Agent | Name | Role |
|---|---|---|
pm-lead |
Jaya | Tech lead — spec, delegation, final review |
designer |
Citra | UI/UX — design spec, design system |
backend-dev |
Sthira | API routes, data layer, validation |
frontend-dev |
Cipta | UI components, kanban board, modals |
qa |
Prajna | End-to-end testing, bug reports |
Workflow: Issues → Jaya (spec + delegate) → Citra (design) → Sthira + Cipta (build) → Prajna (test) → Jaya (SHIP / NEEDS WORK / BLOCK). See CLAUDE.md for the full spec and conventions.
A second squad works alongside MiniHire Dev in this same repo, on the Antigravity (Gemini) runtime — proving Multica can run coding and non-coding work in parallel.
| Agent | Name | Role |
|---|---|---|
coordinator |
Nayaka | Squad leader — reads issues, delegates, doesn't do the work itself |
researcher |
Lekha | Research, fact-finding, competitor/market scans |
content-writer |
Aksara | Drafts docs, posts, copy, any written asset |
data-analyst |
Ganita | Analyzes data, computes metrics, reports |
reviewer |
Waskita | Quality gate — APPROVED / NEEDS REVISION / ESCALATE |
Output is saved to outputs/. Scope evolves by issue, not a fixed roadmap — see CLAUDE.md for conventions.
Product & squad design
- Redho Ramadhani — linkedin.com/in/redhoramadhanihamid · github.com/redhoram
Built by
- Dev squad (Jaya, Citra, Sthira, Cipta, Prajna) — Claude via Multica
- Ops squad (Nayaka, Lekha, Aksara, Ganita, Waskita) — Antigravity (Gemini) via Multica
Every commit carries a Co-Authored-By trailer for the agent that made it — the squads are credited as contributors, ownership of the repo stays with the human.