# OpenKairos

> An open-source autonomous AI daemon that continuously watches your codebase, builds persistent memory, flags vulnerabilities, and sends proactive alerts via Telegram without requiring any prompts.

OpenKairos is an open-source, model-agnostic autonomous background daemon inspired by Anthropic's unreleased KAIROS architecture. Unlike traditional AI coding assistants that wait for prompts, OpenKairos runs 24/7, scanning your repository every 5 minutes, building persistent memory about your project, and proactively alerting you to security issues, failing tests, and code quality problems. It supports multiple LLM providers including Anthropic, OpenAI, DeepSeek, Gemini, and local Ollama models, and delivers notifications via Telegram, macOS notifications, or terminal.

- **Proactive Tick Engine:** *Wakes every 5 minutes to scan git diffs, evaluate changes, and take autonomous action — no prompts needed.*
- **3-Layer Persistent Memory:** *Maintains a high-level `MEMORY.md`, structured topic files, and raw append-only logs to retain full project context across sessions.*
- **AutoDream Consolidator:** *After 30+ minutes of idle time, the daemon consolidates scattered observations into compressed, durable project knowledge.*
- **Model Agnostic:** *Auto-detects your LLM provider from environment variables; works with Anthropic, OpenAI, DeepSeek, Gemini, or Ollama locally.*
- **Omni-Channel Notifications:** *Sends morning briefings, real-time security alerts, and task completion updates via Telegram, macOS notifications, or terminal output.*
- **Two Operating Modes:** *Run in continuous Observation Mode (`kairos watch`) for background monitoring, or Task Mode (`kairos task "..."`) to assign a specific objective the daemon pursues autonomously.*
- **GitHub Webhook Integration:** *Built-in webhook server captures push events, pull requests, and CI/CD pipeline statuses for deep audit reflexes on `git diff` outputs.*
- **MCP Integration:** *Natively integrates with the Model Context Protocol, enabling autonomous connection to calendars, email, CI pipelines, and cloud infrastructure without manual prompts.*
- **Security Auditing:** *Detects leaked API keys, SQL injection vulnerabilities, and risky code patterns before commits are made.*
- **CLI Reference:** *Install via `pip install openkairos`, then use commands like `kairos watch`, `kairos task`, `kairos brief`, `kairos dream`, and `kairos doctor` to manage the daemon.*

## Features
- Autonomous background daemon with 5-minute tick engine
- 3-layer persistent memory (MEMORY.md, topic files, raw logs)
- AutoDream idle consolidation cycle
- Model-agnostic: Anthropic, OpenAI, DeepSeek, Gemini, Ollama
- Telegram, macOS, and terminal notifications
- Observation Mode and Task Mode
- GitHub webhook server for push/PR/CI events
- MCP (Model Context Protocol) integration
- Security vulnerability detection (SQL injection, leaked API keys)
- Morning and evening briefings
- Automatic PR creation for fixes
- 15-second blocking budget for bash commands
- CLI with watch, task, brief, dream, doctor, and mcp-server commands
- Integration with Cursor, Claude, and Aider via kairos integrate

## Integrations
Anthropic Claude, OpenAI GPT, DeepSeek, Google Gemini, Ollama, Telegram, GitHub, ngrok, MCP (Model Context Protocol), Cursor, Aider, pytest, CI/CD pipelines

## Platforms
CLI, API, MACOS, LINUX, WINDOWS

## Pricing
Open Source

## Links
- Website: https://pypi.org/project/openkairos
- Documentation: https://github.com/prabhkesar123/openkairos/blob/main/CONTRIBUTING.md
- Repository: https://github.com/prabhkesar123/openkairos
- EveryDev.ai: https://www.everydev.ai/tools/openkairos