Core Dependencies

The Pieces ecosystem is built on a core infrastructure layer that powers all extensions, plugins, and integrations.

Architecture

graph TB
    POS[PiecesOS]
    POS --> LTM[LTM-2.7 Engine]
    POS --> LLM[LLM Runtime]
    POS --> MCP[MCP Server]
    POS --> API[Local API :39300]
    LLM --> LOCAL[Local Models]
    LLM --> CLOUD[Cloud Models]
    LOCAL --> OLL[Ollama]
    LOCAL --> BUILT[Built-in Models]
    style POS fill:#000,stroke:#fff,color:#fff

PiecesOS

The foundational layer. All Pieces applications communicate with PiecesOS, which handles:

• AI model management (local and cloud)
• Long-Term Memory data collection and storage
• Snippet storage and metadata enrichment
• MCP server for third-party integrations
• Inter-application synchronization

Learn more about PiecesOS

Local AI Models

Pieces supports running LLMs entirely on your device for privacy and offline use. Models can be served via PiecesOS directly or through Ollama integration.

Learn about Local Models

Supported LLMs

Pieces supports 54+ LLMs across multiple providers, both local and cloud-hosted.

See all Supported LLMs

Key Ports

Service	Port	Purpose
PiecesOS API	39300	Main API endpoint
PiecesOS (Linux)	5323	Alternative port on Linux
MCP Server	39300	SSE endpoint for MCP