Duragent

Duragent — A durable, self-contained runtime for AI agents.

Sessions survive crashes. Agents are just files. One binary, zero dependencies.

Use it as a personal AI assistant, or as the foundation for agent-powered products.

What you get	How
Sessions that survive crashes	Append-only event log, attach/detach like tmux
Agents you can read and version	YAML + Markdown — no code required
State you can inspect	Just files on disk — cat, grep, git diff
Deploy anywhere	Single binary, ~10MB, no Python/Node/Docker
Your choice of parts	Swap LLM providers, gateways, and storage backends or bring your own

Quick Start

1. Install and initialize

Download a prebuilt binary from GitHub Releases (Linux x86_64, macOS Intel/ARM), or build from source:

cargo install --git https://github.com/giosakti/duragent.git

duragent init
# Follow the interactive setup

2. Authenticate

export OPENROUTER_API_KEY=your-key  # or: duragent login anthropic

3. Chat with your agent (server starts automatically)

duragent chat --agent <YOUR_AGENT_NAME>

4. Attach to a session later

duragent attach --list       # List attachable sessions
duragent attach SESSION_ID   # Reconnect to existing session

Features

• Durable sessions — crash, restart, reconnect; your conversation survives
• Portable agent format — define agents in YAML + Markdown; inspect, version, and share them
• Memory — agents recall past conversations, remember experiences, and reflect on long-term knowledge
• Tools — bash execution, CLI tools, web search/fetch, scheduled tasks, and background processes, with configurable approval policies
• Skills — modular capabilities defined as Markdown files (Agent Skills standard)
• Context management — token budgeting, history truncation, and priority-based context rendering
• Multiple LLM providers — Anthropic, OpenAI, OpenRouter, Ollama
• Platform gateways — Telegram and Discord via subprocess plugins; group chat with mention gating and debouncing
• HTTP API — REST endpoints with SSE streaming
• Operational tooling — duragent doctor for diagnostics, duragent upgrade for self-update

Modular by Design

Use the built-ins, or swap in your own:

Component	Default	Swappable
Gateways	CLI, HTTP, SSE, Telegram, Discord	Any platform via gateway plugins
LLM	OpenRouter	Anthropic, OpenAI, Ollama, or any provider
Sandbox	Trust mode	bubblewrap, Docker (planned)
Storage	Filesystem	Postgres, S3 (planned)

Workspace Layout

./.duragent/
├── agents/<agent-name>/
│   ├── agent.yaml           # Agent definition (Duragent Format)
│   ├── SOUL.md              # "Who the agent IS" (identity and personality)
│   ├── SYSTEM_PROMPT.md     # "What the agent DOES" (core system prompt)
│   ├── INSTRUCTIONS.md      # Additional runtime instructions (optional)
│   ├── policy.yaml          # Tool execution policy (optional)
│   ├── skills/              # Modular capabilities (SKILL.md files)
│   ├── tools/               # Agent-specific auto-discovered tools
│   └── memory/
│       ├── MEMORY.md        # Curated long-term memory
│       └── daily/           # Daily experience logs
├── sessions/
│   └── <session_id>/
│       ├── events.jsonl     # Append-only event log
│       └── state.json       # Snapshot for fast resume
├── tools/                   # Workspace-level auto-discovered tools
├── schedules/               # Scheduled tasks and run logs
├── processes/               # Background process metadata and logs
└── memory/
    └── world/               # Shared knowledge across all agents

Documentation

Read the Duragent Guide for installation, configuration, and usage.

For contributors: CONTRIBUTING.md | Project Status | Internal Specs

License

MIT