Suzent is a sovereign AI agent: an open-source, local-first agent whose identity, memory, skills, workspace, and runtime stay under your control, independent of any model or platfo

Suzent is a sovereign AI agent: an open-source, local-first agent whose identity, memory, skills, workspace, and runtime stay under your control, independent of any model or platform. For the definition of agent sovereignty and the five-question ownership test, see what makes an agent sovereign.

Getting Started

What is Suzent?: Core concepts and architecture overview
Quickstart: Set up Suzent from scratch in under 5 minutes

Concepts

Core

Providers: Configure model providers (OpenAI, Anthropic, Gemini, Ollama, and more)
Memory: Persistent memory across sessions — how it works and how to configure it
Notebook: Agent-maintained knowledge vault, Obsidian-compatible
Memory Architecture: How memory is written, consolidated, indexed, and read
Memory Internals: Class-by-class reference for contributors

Tools & Capabilities

Tools: Full reference for all built-in tools
Canvas (A2UI): Interactive UI components rendered in the sidebar
Tool Approval: How dangerous tools require user confirmation
Skills: Extend the agent with domain knowledge modules
Filesystem & Sandbox: File access, sandbox execution, and storage paths

Automation & Connectivity

Automation: Cron jobs and heartbeat monitoring
GitHub Sync: Sync portable brain data to a private GitHub repo
Social Messaging: Telegram, Discord, Slack, and Feishu integration
Nodes: Connect and control companion devices remotely
ACP: Serve Suzent as an ACP agent for editors and other clients

Runtime

System Reminders: Runtime context, trust boundaries, and file citations
Retry: Roll back the last agent turn and rerun
Post-Processing: Background tasks after each turn

Development

Stream Recovery Protocol: Snapshot, replay, and persistence acknowledgements
Development Guide: Setup, workflow, production builds, and architecture
Docker Services: Redis, SearXNG, and sandbox configuration
Release Guide: Version bumping and release process
Model Capability Updates: Maintainer workflow for refreshing shipped model metadata

Upgrading

Upgrading: Manual steps when moving an existing install across versions, by release

Documentation