How vstorm-co/agenticos works
AgenticOS is a self-hosted AI teammate platform: users build and govern agents, then work with them in shared chat. The central runtime route reaches agent execution, where configurable capabilities provide tools such as knowledge search, browsing, skills, MCP, code execution, and approvals. RAG ingestion and vector search form a distinct knowledge lifecycle; background jobs support durable work. The README documents these product branches, while the sampled source is incomplete, so unverified connections are shown as dashed with no evidence path.
Main entry points: factory.py Agent runtime · ingestion.py Document ingestion · main.py FastAPI application · v1 Product API
Components 19
Product entry
- Web console
frontend/src - FastAPI application
backend/app/main.py - Product API
backend/app/api/routes/v1
Agent runtime
- Agent runtime
backend/app/agents/factory.py - Agent capabilities
backend/app/agents/capabilities - Model providers
Knowledge lifecycle
- Document ingestion
backend/app/services/rag/ingestion.py - Sync workers
backend/app/worker/tasks/rag_tasks.py - Vector search
backend/app/services/rag/vectorstore.py - RAG settings
backend/app/services/rag/models.py - File storage
backend/app/services/file_storage.py
Governance and state
- Application database
backend/app/db - Background jobs
backend/app/core/background.py - Access controls
backend/app/core/permissions.py - Audit trail
backend/app/core/audit.py
Channels and integrations
- Team channels
backend/app/services/channels - MCP integrations
backend/app/agents/mcp.py - Coding agents
backend/app/agents/capabilities/coding_agent
Other components
- Team user
Connections 21
- Team user → Web console: uses
README.md - Web console → FastAPI application: sends requests
- FastAPI application → Product API: mounts routes
backend/app/main.py - Product API → Agent runtime: starts agent runs
- Agent runtime → Agent capabilities: uses capabilities
README.md - Agent runtime → Model providers: calls models
README.md - Team user → Agent runtime: chats with
README.md - Agent capabilities → Document ingestion: searches knowledge
README.md - Sync workers → Document ingestion: processes documents
backend/app/worker/tasks/rag_tasks.py - Document ingestion → Vector search: indexes documents
backend/app/services/rag/ingestion.py - Vector search → RAG settings: uses settings
backend/app/services/rag/vectorstore.py - Document ingestion → File storage: ingests files
- Sync workers → Application database: reads document state
backend/app/worker/tasks/rag_tasks.py - Product API → Access controls: checks access
backend/app/api/routes/v1/_sharing_routes.py - Agent runtime → Audit trail: records activity
- FastAPI application → Background jobs: starts background work
backend/app/main.py - Team user → Team channels: works through
README.md - Team channels → Agent runtime: routes messages
README.md - Agent runtime → MCP integrations: uses MCP
README.md - Agent runtime → Coding agents: delegates coding
README.md - Agent runtime → Application database: persists runs