EN·ES·DE·PT·FR
⌘K

AIdaemon

A personal AI agent that runs as a daemon. Always on, always learning. Chat from Telegram, extend with MCP, powered by any LLM.

aidaemon is a self-hosted AI agent written in Rust that runs as a background service on your machine. It connects to any OpenAI-compatible LLM provider, communicates via Telegram, Slack, or Discord, and can execute tools, manage its own configuration, remember facts, browse the web, and spawn sub-agents — all autonomously.

Key Features

  • Daemon architecture — runs as systemd/launchd service, always available
  • Multi-channel — chat via Telegram, Slack, or Discord, multi-user access control
  • Agentic tool use — autonomous multi-step reasoning with up to 10 iterations
  • MCP integration — extend with any Model Context Protocol server
  • Persistent memory — SQLite-backed history with semantic search via embeddings
  • Multi-model routing — automatic Fast/Primary/Smart tier selection
  • Email triggers — IMAP IDLE monitoring for inbox notifications
  • Command approval — interactive Telegram approval for shell commands
  • Skills system — dynamic prompt enhancement via markdown files
  • Self-maintenance — reads, updates, validates, and restores its own config
  • Browser automation — Chrome with persistent login sessions, screenshots, and form filling
  • Web research — built-in web search (DuckDuckGo/Brave) and URL fetching
  • Sub-agent spawning — recursive agent delegation for complex tasks
  • CLI agent delegation — delegate to Claude, Gemini, Codex, Aider, etc.
  • Scheduled tasks — cron-style recurring tasks with natural language parsing
  • File transfer — send and receive files via Telegram with path security
  • Secrets management — OS keychain and environment variable support
  • Token cost tracking — per-model usage stats, daily budgets, /cost command
  • Event sourcing — immutable event log with daily consolidation into facts and procedures
  • Goals + tasks — break bigger work into trackable tasks and keep progress moving
  • Health monitoring — HTTP, TCP, command, and file probes with alerting
  • Dynamic skills — install from registries or auto-promote repeated procedures
  • Self-updater — auto-update from GitHub releases with configurable modes
  • Discord integration — slash commands, interactive approval buttons, multi-bot support
  • Command risk assessment — 4-level risk scoring (Safe/Medium/High/Critical) for terminal commands
  • People intelligence — a personal contact book that remembers birthdays, preferences, and relationships for you

Architecture at a Glance

architecture
Telegram ──┐
Slack    ──┤──> ChannelHub ──> Agent ──> LLM Provider
Discord  ──┘         │
                     ├──> Tools
                     │    ├── terminal (risk assessment)
                     │    ├── system info
                     │    ├── memory (facts)
                     │    ├── config manager
                     │    ├── web search + fetch
                     │    ├── browser (optional)
                     │    ├── file transfer
                     │    ├── sub-agents + CLI agents
                     │    ├── health probes
                     │    ├── manage skills
                     │    ├── manage people
                     │    ├── scheduler
                     │    └── MCP tools (dynamic)
                     │
                     ├──> Events
                     │    ├── immutable event log
                     │    └── daily consolidation
                     │
                     ├──> Goals + Tasks
                     │    ├── break work into tasks
                     │    └── track progress + retries
                     │
                     ├──> State
                     │    ├── SQLite persistence
                     │    └── in-memory working mem
                     │
                     └──> Facts
                          └── injected into system prompt

Triggers ───> EventBus ───> Agent
   └── IMAP IDLE

Health ───> GET /health (axum)

Quick Links

NextOverview