# Mendry

> Self-hosted AI incident investigation. Installation is source evaluation (Preview), not supported production deployment.

## Try with your existing bot

- [Try with your bot](https://www.mendry.net/docs/get-started/bot/)
- [用已有 Bot 体验](https://www.mendry.net/zh-cn/docs/get-started/bot/)
- [Self-contained Mendry Bot Skill](https://www.mendry.net/mendry-bot.md)

The shared Skill responds in the user's language. Grok, Meta Muse.ai, and Claude cloud sessions have different setup paths. Verify tools and triggers before claiming automatic operation; start manually when event delivery is unavailable. Deploying Mendry is a separate choice, not a prerequisite. Live-account end-to-end validation remains pending.

## Install with an AI assistant

Follow the user's explicit language preference, otherwise their current conversation language. For Chinese use the Chinese runbook; for English use the English runbook. If unclear or another language is used, ask Chinese or English. Keep commands and identifiers unchanged; use the selected language for consent, troubleshooting, and handoff. Confirm the target and permissions before executing. Never overwrite existing configuration or encryption keys. Do not rerun completed steps when switching languages.

- [English installation entry](https://www.mendry.net/docs/get-started/ai-install/)
- [中文安装入口](https://www.mendry.net/zh-cn/docs/get-started/ai-install/)
- [English executable runbook](https://www.mendry.net/agent-install.en.md)
- [中文可执行安装手册](https://www.mendry.net/agent-install.md)
- [Source repository](https://github.com/Funovate/Mendry)

The runbooks are generated from the same source revision as this documentation build. After cloning, use docs/agent-install.md or docs/agent-install.en.md from the checked-out revision. Stop on disagreements with the checkout. Model calls and remote access require separate consent. A successful health check is not proof of end-to-end AI analysis.