Open-source self-hosted AI Agent platform featuring voice-driven workflows, episodic memory, multi-view orchestration, and enterprise-grade integrations. Users can drive complex automation via natural language while keeping data entirely in local environments.
Atom is a self-hosted AI Agent automation platform built with Python and Next.js, designed to drive complex business processes through voice or natural language interaction.
Core Capabilities#
Voice Interface & Visual Orchestration#
- Build complex workflows via voice without proprietary syntax
- Canvas provides rich interactive demos (charts, forms, Markdown)
- Multi-view synchronized orchestration: browser, terminal, canvas real-time sync
- Real-time reasoning visualization with intelligent error solutions
Specialized Agent System#
- Pre-built role templates: Sales Agent (CRM, lead scoring), Marketing Agent (campaigns, social media), Engineering Agent (PR notifications, deployments)
- Four-tier maturity governance: Student → Restricted → Supervised → Autonomous
- Graduation verification: Agents must pass validation for elevated permissions
Episodic Memory & Retrieval#
- Experience-based learning system
- Hybrid storage: PostgreSQL (Enterprise) / SQLite (Personal) + LanceDB vector database
- Four retrieval modes: temporal, semantic, sequential, contextual
- FastEmbed local embedding generation (10-20ms latency)
Integrations & Extensions#
- Claims support for 46+ commercial integrations: Slack, Gmail, HubSpot, Salesforce, Zendesk
- Compatible with 5,000+ OpenClaw/ClawHub community skills
- Python (350K+ PyPI packages) and npm (2M+ packages) ecosystem support
- CDP-based browser automation (scraping, form filling)
Deployment#
Recommended: Docker (~5 minutes)
git clone https://github.com/rush86999/atom.git
cd atom
cp .env.personal .env
docker-compose -f docker-compose-personal.yml up -d
One-click Scripts: ./start.sh (macOS/Linux), start.bat (Windows)
Native Install: Configure backend (Python + FastAPI) and frontend-nextjs separately
Access: http://localhost:3000
Configuration#
Edit .env file to configure LLM API keys:
OPENAI_API_KEYorANTHROPIC_API_KEYorGEMINI_API_KEY
Ports: Backend default 8000, Frontend default 3000 (auto-increment on conflict)
Editions:
- Personal Edition: Free, single user, SQLite
- Enterprise Edition: Multi-user, PostgreSQL, monitoring included
Typical Use Cases#
| Department | Example Scenario |
|---|---|
| Sales | HubSpot new lead → Research → Score → Slack notification |
| Finance | Gmail PDF invoice → Extract → Match QuickBooks → Flag discrepancies |
| Support | Zendesk ticket → Sentiment analysis → Urgent routing → Draft response |
| HR | BambooHR new employee → Configure → Invite → Schedule onboarding |
Security Features#
- BYOK (Bring Your Own Key): Data never leaves user environment
- Data at rest encryption, supply chain protection
- Enterprise skill scanning (21+ malicious pattern detection)
- Complete audit logging of all Agent operations
Important Notes#
⚠️ The pip install atom-os mentioned in README is confirmed non-existent on PyPI. Please use Git Clone or Docker deployment.
⚠️ Official website https://www.atomiclife.app is currently inaccessible; GitHub repository is the only reliable source.