What is AI Butler?

AI Butler is the open-source, self-hosted personal AI agent you own end-to-end. One Go binary. No dependencies. Runs offline. Remembers you across every conversation, every channel, every week. Works with Claude, GPT, Gemini, or your own local Ollama — zero cloud lock-in.

It’s what you get when you take ChatGPT, strip out the surveillance, add a real memory system, plug in 12 messaging apps, and let the agent actually do things on your machine through a capability-gated tool registry.

What makes it different

Commercial AI chat (ChatGPT, Claude.ai, Gemini)	AI Butler
Lives in the cloud	Lives on your machine
Forgets you between sessions	Remembers you forever (FTS5 + knowledge graph + vector embeddings)
One model, no choice	Any model — Claude, GPT, Gemini, Grok, Ollama, LM Studio, vLLM, Groq, DeepSeek
Chat-only	12 channels — web, terminal, Telegram, WhatsApp, Slack, Discord, Teams, Google Chat, LINE, IRC, webhook, Nostr
Pay monthly, cancel = gone	Free, Apache 2.0 — you pay your model provider directly, or run fully local for $0
Can’t access your files	Filesystem-aware via MCP — reads your actual workspace
Black box	Capability-gated + audit-logged — every tool call is visible and permissioned
Can’t automate	Natural-language scheduling — “every weekday at 8am brief me” → cron

The short pitch

You install one binary. You open a web chat. You tell it who you are, what you’re working on, what you want it to remember. From that moment on:

• Every future conversation draws on that memory — on the web, in Telegram, on the terminal, wherever
• Every scheduled task you set up runs on its own and delivers to whichever channel you pick
• Every file, tool, and connected app is available through natural language — no APIs to learn
• Every dollar spent is tracked, capped, and visible in a live budget meter

Your data stays on your laptop. Your models stay in your control. Your agent stays yours.

Works everywhere you chat

12 channels wired up — web chat ready today, 11 others in beta. Telegram, WhatsApp, Slack, Discord, Teams, Google Chat, LINE, IRC, custom webhook, Nostr, plus the built-in web chat and terminal REPL. Same agent, same memory, every platform. Start a conversation on your laptop, continue it from your phone on your commute, finish it from a terminal SSH session that evening — all one thread to the agent, all drawing on one memory store.

See Channels for the full list and setup.

Memory that actually works

This is the #1 differentiator. Three memory systems running in parallel, fused with Reciprocal Rank Fusion:

1. FTS5 full-text search — SQLite’s built-in BM25 keyword index over every message, note, and extracted fact
2. Knowledge graph — entities (people, projects, places, decisions) automatically extracted from conversation, stored relationally in SQLite, queryable with recursive CTEs
3. Vector embeddings — semantic similarity via Ollama’s nomic-embed-text (default, local) or any OpenAI-compatible embedding provider

Ask about something you mentioned weeks ago — it finds it, even if you don’t remember the exact words. See the Memory feature page for the architecture.

Security first

AI Butler is built with a security-first architecture. 59 internal security audit passes. RBAC with admin/user/viewer/agent roles. FIDO2/WebAuthn + OIDC SSO + TOTP. Capability-gated tools with tamper-evident audit trail. WASM plugin sandbox (Extism). SSRF protection. Shell command allowlisting. Zero CVEs (govulncheck verified).

Details: Security Model and Security Audit.

Your data, your machine

Self-hosted. Single Go binary. Runs on Raspberry Pi, laptop, VPS, or Kubernetes. Apache 2.0 licensed with explicit patent grant. Zero telemetry unless you opt in. Everything in a local SQLite file you can back up with cp.

Quick install (3 commands)

git clone https://github.com/LumabyteCo/aibutler.git
cd aibutler && CGO_ENABLED=0 go build -o aibutler .
./aibutler vault set anthropic_api_key sk-ant-... && ./aibutler run

Open http://localhost:3377 — you’re chatting with your own personal AI in under a minute.

Using Ollama for fully-local? Skip the vault set line — AI Butler auto-detects Ollama on startup.

Next steps

• Quick Start with screenshots → — 5-minute walkthrough from git clone to your first real conversation, plus 5 live examples showing memory + scheduling + file exploration in action
• Choose your AI model → — compare Claude / GPT / Gemini / Ollama and their cost/quality tradeoffs
• Install — full install options (source, release binary, Docker, Kubernetes, systemd, Raspberry Pi)
• Integrate an MCP server → — extend AI Butler with any of the 58+ platforms in the MCP ecosystem
• FAQ → — the 25 most common questions about AI Butler