Create AI teams that work together — powered by real intelligence, not templates.
Describe what you need → openminions designs your team → agents execute with full transparency
openminions is a multi-agent orchestration framework built on top of @techwavedev/agi-agent-kit. It creates teams of specialized AI agents that collaborate on complex tasks — from content creation to code reviews, competitive analysis to email campaigns.
Unlike template-based tools, openminions uses Qdrant-backed semantic search to automatically discover and wire the best skills for your task. Your agents aren't pre-built — they're designed on the fly based on your intent.
A team (internally called a "squad") is a group of AI agents that collaborate on a task. Each agent has a specific role and set of tools. They run in a pipeline with checkpoints where execution pauses for your approval before continuing.
Example — Content Pipeline:
📋 Researcher → ✍️ Writer → 🎨 Designer → ✅ Editor
↑ ↑
checkpoint checkpoint
"Review research" "Approve final draft"
- Researcher gathers information from web sources
- Writer produces compelling long-form content
- Designer generates hero images and illustrations
- Editor reviews, polishes, and SEO-optimizes the piece
Each agent transitions through states: idle → working → delivering → done — all visible in the real-time dashboard.
- Solo developers — automate research, documentation, and code review pipelines
- Content creators — produce blog posts, social media campaigns, and newsletters with AI teams
- Agencies & freelancers — create reusable pipelines for client work
- Marketing teams — generate consistent content with human approval at key stages
- Engineering teams — automate code analysis, security audits, and documentation generation
| Category | Examples |
|---|---|
| 📝 Content production | Blog posts, social media campaigns, newsletters, documentation |
| 🔍 Research & analysis | Competitive analysis, market research, trend monitoring |
| 💻 Code operations | Code review, security audit, documentation generation, refactoring plans |
| 📧 Outreach | Personalized email sequences, lead research, campaign management |
| 📊 Data → Insights | Transform raw data into reports, presentations, and dashboards |
| 🎨 Visual content | Generate images, design social media assets, create brand materials |
Prerequisites: Node.js 18+ · Python 3.10+ · agi-agent-kit
git clone https://github.com/techwavedev/openminions.git
cd openminions
npm install
cd ui && npm install && cd ..node bin/cli.js initThe setup wizard walks you through:
- Language — English, Português, Español
- IDE — Antigravity, Claude Code, Cursor, Copilot, OpenCode, Codex
- Scenario — pick from 6 predefined templates or describe your own
- Review — see the team lineup and approve before anything runs
# From a scenario
python3 bin/runner.py --squad data/squads/blog-content-pipeline
# From natural language
python3 bin/runner.py --intent "Research AI trends and write a detailed report" --auto
# Preview without execution
python3 bin/runner.py --squad data/squads/blog-content-pipeline --dry-runnpm run dev
# Open http://localhost:5173The Phaser-powered dashboard shows your agents working as animated sprites, with live status transitions and step tracking.
| IDE | Status |
|---|---|
| Antigravity (Gemini) | ✅ Available |
| Claude Code | ✅ Available |
| Cursor | ✅ Available |
| VS Code + Copilot | ✅ Available |
| OpenCode | ✅ Available |
| Codex (OpenAI) | ✅ Available |
These are ready-to-use team templates. Pick one during init or create your own:
| Scenario | Agents | Description |
|---|---|---|
| ✍️ Blog Content Pipeline | Researcher → Writer → Designer → Editor | Research, write, illustrate, and polish blog posts |
| 🔍 Code Review Squad | Analyzer → SecurityAuditor → Optimizer → Reporter | Full codebase analysis with security and performance |
| 📱 Social Media Campaign | Strategist → Copywriter → VisualDesigner → Scheduler | Multi-platform social campaign with visuals |
| 📊 Competitive Analysis | Scout → PricingAnalyst → FeatureMapper → Strategist | Deep competitor research with strategic recommendations |
| 📧 Email Outreach | LeadResearcher → Copywriter → Reviewer → Sender | Prospect research, email crafting, and delivery via Resend |
| 📚 Documentation Generator | CodeScanner → APIDocWriter → GuideWriter → Formatter | Automated documentation from codebase analysis |
| Command | What it does |
|---|---|
node bin/cli.js init |
Interactive setup (language, IDE, scenario) |
node bin/cli.js scenarios |
List all available scenarios |
node bin/cli.js teams |
List created teams and their status |
node bin/cli.js run --intent "..." |
Design + execute a team from natural language |
node bin/cli.js run --squad <path> |
Execute an existing team |
node bin/cli.js dashboard |
Start the visual dashboard |
python3 bin/architect_wizard.py --intent "..." |
Design a team with Atlas (the architect) |
openminions follows a 3-layer architecture: Directives → Orchestration → Execution.
┌─────────────────────────────────────────────────────────┐
│ openminions │
│ │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌────────┐ │
│ │ CLI │ │ Architect│ │ Runner │ │ Dash │ │
│ │ (init) │→ │ (Atlas) │→ │(execute) │→ │ (view) │ │
│ └──────────┘ └──────────┘ └──────────┘ └────────┘ │
│ ↕ ↕ ↕ ↕ │
│ ┌──────────────────────────────────────────────────┐ │
│ │ Qdrant Memory Layer │ │
│ │ Skills index · Squad history · Decisions │ │
│ └──────────────────────────────────────────────────┘ │
│ ↕ │
│ ┌──────────────────────────────────────────────────┐ │
│ │ @techwavedev/agi-agent-kit │ │
│ │ Skills · Execution scripts · Local LLM │ │
│ └──────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────┘
What makes it different:
- 🧠 Intelligence, not templates — skills are discovered via semantic search, not manually installed
- ⚡ Dynamic teams — agents are created on-the-fly from your intent, not predefined
- 🔒 Validation gates — security checks before and after every execution step
- 👁️ Full transparency —
runs.mdandmemories.mdlog everything in human-readable format - 🌍 Multi-IDE, multi-language — works across 6 IDEs and 3 languages out of the box
openminions is free and open source as software. You can run it at zero cost with:
- Antigravity (Gemini free tier)
- Local LLMs via Ollama (Gemma, GLM, etc.)
If using paid stacks (Claude Code, OpenAI API), each team execution consumes tokens. The cost depends on:
- Number of agents in the pipeline
- Complexity of the task
- Model used for each step
💡 Use
--dry-runto preview the full pipeline without consuming any tokens.
This project takes security seriously:
- Pre-push hook —
bin/security_gate.pyscans for secrets before every push - GitHub Secret Scanning — enabled with push protection
- CodeQL Analysis — runs on every PR (JavaScript/TypeScript + Python)
- TruffleHog — deep secret scanning in CI
- Dependency Review — blocks PRs introducing known vulnerabilities
- Dependabot — automatic security updates
- Branch Protection — all changes require PR with passing checks
- Squash Merge Only — clean commit history, no accidental leaks
See SECURITY.md for the full security policy and vulnerability reporting.
See ROADMAP.md for the full development plan from v0.1.0 to v1.0.0.
openminions is built and maintained by Elton Machado as the first product evolution from the @techwavedev/agi-agent-kit framework.
The project was born from a real need to orchestrate AI agents intelligently — not with static templates, but with semantic understanding that adapts to each task.
Contributions are welcome via Issues and Pull Requests. See the issue templates before opening anything.
Apache-2.0 — see LICENSE for details.