open source · security first · runs locally
Kodo
Security-first AI agent framework _
An open-source agent that runs on your hardware, communicates through your messaging platforms, and keeps your data under your control.
01
Heartbeat Loop
Not a chatbot. A continuous agent loop that pulses on interval β checking messages, running tasks, taking autonomous action even when idle.
02
Security First
Capability-based permissions, sandboxed skill execution, input sanitization, and encrypted memory at rest. Your agent, locked down by default.
03
Multi-Provider LLM
Claude, GPT, Gemini, Ollama, and 10+ more. Switch models with one line of config. No vendor lock-in.
04
Lightweight & Portable
Minimal dependencies, no bloated frameworks. Runs on a laptop, a Raspberry Pi, or a server. Your hardware, your rules.
$ gem install kodo-bot
$ kodo init
$ export ANTHROPIC_API_KEY="sk-ant-..."
$ export TELEGRAM_BOT_TOKEN="123456:ABC-..."
$ kodo start
▸ Kodo v0.1.0 starting...
▸ Channel: Telegram (connected)
▸ LLM: Claude via Anthropic
▸ Heartbeat: 60s interval
▸ Listening on http://localhost:7377