Your AI operating system for development — a desktop coding assistant with multi-agent orchestration, browser automation, terminal integration, and deep code intelligence.
- Multi-agent AI engine — Manager, Coder, Research, Browser, QA, and Memory agents collaborate on complex tasks with streaming responses, multi-model support (OpenAI, Anthropic, MCP), and tool orchestration.
- Code intelligence — Symbol indexing across 8+ languages, dependency graph visualization, semantic code search, and architecture-aware project mapping.
- Browser workspace — Headless browser automation via CDP with an embedded live viewport, console/network inspector, annotation tools, and session persistence.
- Terminal & execution — Full PTY terminal emulation with real-time command streaming, approval gates, command allowlists, and sandboxing.
- Memory & context — Tiered memory system (ephemeral, session, project, global scopes) with intelligent context compression and workspace-aware relevance.
git clone https://github.com/agenticos/agenticos.git
cd agenticos
npm install
npm run dev- Add an AI provider in Settings → Providers
- Open a workspace folder
- Start coding — ask the AI to explain code, refactor, or build features
npm run cli -- --print "Explain the event system"
echo "Hello" | npm run cli -- --stdin
npm run cli -- --file prompt.txt --jsonsrc/
├── main/ # Electron main process (IPC, window management, services)
├── preload/ # Context bridge (preload script)
└── renderer/ # React application
├── core/ # Kernel, routing, error boundaries
├── components/ # UI components organized by domain
├── pages/ # Route pages
├── runtime/ # AI execution engine (agents, tools, memory, context, streaming)
├── stores/ # Zustand state stores
├── lib/ # Utilities and services
└── types/ # TypeScript type definitions
packages/
├── providers/ # Provider transport layer
├── shared/ # Shared types and utilities
└── ui/ # Shared UI components
tests/ # Test files organized by domain
User Input → RuntimeOS → AgentExecutor → Tools → File System / Browser / Terminal
↓ ↓
ExecutionSessionManager ← ExecutionEvent protocol
↓
StreamManager → TimelineStore → React UI
| Category | Technology |
|---|---|
| Desktop | Electron 42 |
| UI | React 19, Tailwind CSS 4, Radix UI, Framer Motion |
| Language | TypeScript 6 (strict mode) |
| State | Zustand |
| Editor | Monaco Editor |
| Terminal | xterm.js + node-pty |
| Browser | Playwright Core (CDP) |
| Testing | Vitest 3 |
| Build | Vite 6 + electron-vite + electron-builder |
| AI | Multi-provider (OpenAI, Anthropic, MCP) |
| Command | Description |
|---|---|
npm run dev |
Development mode with hot reload |
npm run build |
TypeScript check + production build |
npm test |
Run all tests (vitest) |
npm run typecheck |
TypeScript strict type checking |
npm run lint |
ESLint code quality |
npm run dist |
Build distribution packages |
npm run cli |
Headless CLI mode |
880+ passing tests across 68 test files.
npm test # Full suite
npx vitest run tests/agent-system # Agent tests
npx vitest run tests/reliability # Reliability tests
npx vitest run tests/browser # Browser tests
npx vitest run tests/memory # Memory tests| Operation | p50 | p95 | p99 |
|---|---|---|---|
| Search (1k files) | 0.4ms | 5.1ms | 17.3ms |
| Search (10k files) | 2.7ms | 10.7ms | 83.5ms |
| Search (50k files) | 7.3ms | 22.9ms | 30.6ms |
Production Readiness Score: 82%
| Category | Score |
|---|---|
| Agent System | 93% |
| Reliability | 90% |
| Architecture | 85% |
| Code Intelligence | 85% |
| UX | 86% |
| Security | 60% |
| Scalability | 62% |
See CONTRIBUTING.md for guidelines on code of conduct, development workflow, commit conventions, and pull request process.
MIT — see LICENSE.
Built with Electron, React, TypeScript, and the open-source community.