Skip to content

README.md

Latest commit

 

History

History
129 lines (85 loc) · 5.32 KB

README.md

File metadata and controls

129 lines (85 loc) · 5.32 KB

DEVS Logo

DEVS

Delegate complex tasks to a swarm of AI agents
through a universally accessible, privacy-first platform.

[ https://devs.new ]

Features • Self-Hosting • Contributing • Documentation

Features

DEVS is an open-source, browser-native platform that enables users to delegate complex tasks to a swarm of AI agents. Here are some of its standout features:

Artificial Intelligence

  • 🤖 LLM provider independence: BYOK support for OpenAI, Anthropic, Google Gemini, Mistral, OpenRouter, HuggingFace, Ollama, LM Studio, Vertex AI, and more.
  • 🧠 Local AI models: Run HuggingFace open models directly in your browser via WebGPU.
  • 💰 Traces & Cost tracking: LLM observability with real-time cost tracking and performance metrics.
  • 🖥️ Sandboxed code execution: Agents can write and execute JavaScript (QuickJS) and Python (Pyodide) in WASM-isolated sandboxes — no server needed.

Swarm Intelligence

  • 🤖 AI Studio: Pre-built agents, custom agent builder with AI-generated portraits.
  • 📝 Methodologies: 8D, A3, Agile, AOSTC, DMAIC, PDCA, Scrum, YOLO, and your own.
  • 🧠 Agent Memory: Agents learn and remember information from conversations with human review.
  • 🔧 Tools: Equip agents with capabilities like Wikipedia search, arXiv search, calculator, code execution, and more.
  • 🎯 Multi-agent orchestration: Automated task coordination with decomposition, dependency resolution, and parallel execution.
  • 🔌 Skills: Installable capability bundles that extend what agents can do.
  • 🎭 Hyper meta-prompting: Multi-layered prompt generation for dynamic task handling.
  • 👥 Dynamic team formation: Mimicking human organizational structures.

Integrations

  • 🔗 Connectors: Integrations with Google Drive, Gmail, Calendar, Notion, Slack, Figma, Dropbox, OneDrive, Outlook, and more.
  • 🏪 Marketplace: Extensible platform with apps, agents, connectors, and tools.
  • 🎤 Meeting Bot: AI agents join Google Meet as real participants.

Privacy

  • 🌐 Browser-native: The browser is the OS. Runs entirely in your browser, no server dependencies.
  • 🛡️ Privacy-first: All data stays on your device, no tracking, no cookies.
  • 🥷 Offline capable: Works without Internet connection after initial load.
  • 💾 Local Backup: Preserve your data with bidirectional sync to a local folder on your host, with human-readable Markdown files.

User Experience

  • 📱 Mobile-first design: Optimized for touch and small screens.
  • 🌐 Multi-language support: English, French, German, Spanish, Arabic, Korean.
  • 🔍 Global Search: Unified search across agents, conversations, tasks, and files (Cmd/Ctrl+K).
  • 💄 Customizable: Tailor the platform to your specific needs.
  • 📂 Spaces: Multi-workspace isolation for different projects or contexts.

Collaboration

  • 🌐 Universal access: Web-based platform accessible from anywhere: https://devs.new
  • 🔄 P2P Sync: Cross-device synchronization with CRDT-based conflict resolution.
  • 🤝 Team collaboration: Peer-to-peer networking and sharing.
  • 📱 QR Code support: Easily share and access your platform configuration.

Deployment

  • ⚙️ Open-source: Community-driven development and transparency.
  • 🐳 Docker-ready: One-command self-hosting with an image of ~60MB (registry).

Self-Hosting

🐳 Docker

 
docker run -d -p 80:80 codename/devs

Then open http://localhost in your browser.

🐳 Docker compose

Alternatively, check out the compose.yaml file for a Docker Compose setup.

services:
  devs:
    image: codename/devs
    ports:
      - 8080:80

See https://hub.docker.com/r/codename/devs for more details.

Documentation

Detailed documentation is available in the docs/ directory:

  • Architecture — System architecture and data layer
  • Conventions — Code style, naming, and patterns
  • Decisions — Architectural decision records
  • Glossary — Term definitions
  • Vision — Project vision and design principles
  • TODO — Feature status and roadmap

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for details.

License

This project is licensed under the MIT License. See the LICENSE file for details.