Ask your texts real questions, recover buried details, and turn vague plans into to-dos and calendar events โ right from your Mac.
Download Bert for MacRequires macOS 13.2+ · Free
"This does exactly what I want Siri to do for my messages!"
Bert runs entirely on your Mac. Your messages are indexed locally in an embedded database โ nothing is uploaded, stored remotely, or shared. You bring your own API key for AI features.
Download Bert and move grant Full Disk Access. Bert indexes your messages locally with full-text and vector search.
Add your API key and hit Sync. Bert sends the conversations directly to your LLM provider to summarize and find to-dos and calendar events.
Search your history, review action items, and never let an important message slip through the cracks.
No. Bert can't send, delete, or modify any of your messages. That's your job. Bert is read-only โ it just helps you stay organized and gives you a nudge when something needs your attention.
No. Bert sends the conversations directly to your LLM provider to summarize and find to-dos and calendar events. Your messages are never stored on a server.
Bert only connects to two services: the Claude API (Anthropic) or the OpenAI API โ and only when you use AI features like summaries and action items. These calls use your own API key and are limited to summarizing and finding messages. Bert cannot access any other websites, services, or third-party tools. Your messages stay between your Mac and the AI provider you choose.
Bert includes an MCP server that lets you connect it to Claude Code or other MCP-compatible tools. If you use this, be aware that the connected service will be able to query your messages through Bert โ just as it can already access anything else on your disk. This isn't unique to Bert: any tool with Full Disk Access (or any MCP client you grant access to) can read your messages. Use the same caution you would with any provider that has access to your machine.
Bert is open source, licensed under GPL v3. Check out the repo, file issues, or contribute.
View on GitHub