written in Rust · MIT License

aidaemon — Open Source Personal AI Agent Daemon

Stop doing everything yourself.
Hand off research, reminders, and repetitive tasks to an assistant that's always ready.
Reach it from Telegram, Slack, or Discord without being at your desk.

View on GitHub $ cargo install

~/aidaemon

$ cargo install aidaemon
Installing aidaemon v0.3.0
Installed to ~/.cargo/bin/aidaemon

$ aidaemon
✓ Config loaded from config.toml
✓ SQLite state store initialized
✓ Event store initialized
✓ MCP tools discovered: 12 tools from 3 servers
✓ Telegram bot connected
✓ Discord gateway connected
◆ daemon active — listening on all channels
$

Native + OpenAI-compatible providers

Anthropic (native)

Google Gemini (native)

DeepSeek

OpenAI

OpenRouter

Ollama

Features

What You Can Do with an AI Agent Daemon

An assistant that actually gets things done, not just answers questions.

Access from anywhere

Message it from Telegram, Slack, or Discord. Share access with your family or team.

telegram + slack + discord

Handles complex tasks

Give it a goal and walk away. It creates a persistent plan, tracks each step, and recovers from crashes.

plans + agent loop

Connects to your tools

Link your files, calendars, and services so it can help with real work.

extensible

Never forgets

Every action is recorded as an immutable event. Facts and procedures are extracted daily. Your data stays private and encrypted.

event sourcing + memory

Browses the web for you

Check websites, fill out forms, grab screenshots. All without lifting a finger.

chromium

Tackles big projects

Breaks down complex work into smaller pieces and handles them for you.

recursive

Works with other AI tools

Coordinates with Claude, Gemini, Copilot, and more when needed.

delegation

Researches for you

Searches the web, reads articles, and summarizes what it finds.

search + fetch

Set it and forget it

Schedule tasks with simple phrases like "every morning at 9am" or "next Tuesday."

scheduler

You stay in control

Every command is risk-scored (Safe → Critical). It asks before running anything sensitive. Approve or deny with one tap.

risk assessment + approval

Monitors your services

Define health probes for your services — HTTP, TCP, command, or file checks — and get alerted when something breaks.

health probes

Learns new abilities

Install skills from registries or let it auto-promote repeated procedures into reusable skills.

dynamic skills

Keeps itself current

Checks GitHub for new releases and updates itself automatically — or asks first, your choice.

self-updater

Always on, always ready

Starts automatically with your computer. No need to launch anything.

daemon

Architecture

How aidaemon Works

A simple flow: you send a message, it thinks and takes action, you get results. It can handle multiple conversations at once without slowing down.

Telegram ──┐ Slack ──┼──> Agent ───> LLM Provider Discord ──┘ │ ├──> Tools │ ├── terminal (risk assessment) │ ├── browser (headless) │ ├── web search + fetch │ ├── sub-agents + CLI │ ├── health probes │ └── MCP tools │ ├──> Events │ ├── immutable event log │ └── daily consolidation │ ├──> Plans │ ├── detect + generate │ └── track + recover │ ├──> State │ ├── SQLite + vectors │ └── working memory │ └──> Memory └── facts + recall Dashboard ──> /dashboard Health ──> /health

Message arrives

A message from Telegram, Slack, Discord, or a scheduled trigger enters via async channels.

Plan detection

The Agent checks if the request needs a multi-step plan. If so, it generates steps and tracks progress.

Agent loop begins

The Agent loads event context and working memory, injects long-term facts, and sends to the LLM.

Tools are called

The Agent executes tool calls — terminal (with risk assessment), browser, web research, sub-agents, or MCP tools.

Loop continues

Tool results feed back into the LLM for up to 10 iterations, with stall and repetition detection.

Response delivered

The response is sent back. All messages are persisted to SQLite with vector embeddings.

Events recorded

Every action is stored as an immutable event. Daily consolidation extracts facts and procedures.

Quickstart

Install aidaemon in Minutes

Get started in minutes. The setup wizard walks you through configuration step by step.

Install
curl -sSfL https://get.aidaemon.ai | bash
Or: brew install davo20019/tap/aidaemon · cargo install aidaemon

Run the setup wizard
On first run, aidaemon launches an interactive wizard to configure your LLM provider and channel tokens.

Configure MCP servers (optional)
Add MCP server definitions to config.toml to extend with filesystem, databases, and more.

Install as service
aidaemon install-service
Runs as systemd (Linux) or launchd (macOS).

config.toml

[provider]
api_key = "keychain"
base_url = "https://api.openai.com/v1"

[provider.models]
primary = "gpt-4o"
fast = "gpt-4o-mini"
smart = "o1-preview"

[telegram]
bot_token = "keychain"
allowed_user_ids = [123456789]