LangChain: OpenWiki Brains gives agents proactive memory — automatically collects context from Gmail, Notion and Git into Markdown

On July 10, 2026, LangChain introduced OpenWiki Brains, an open-source framework that gives AI agents what it calls “proactive memory.” Memory is one of the biggest limitations of agents: most systems only remember what the user tells them directly, while OpenWiki Brains actively collects context from the sources where the user already works.

How does proactive memory work?

The framework runs locally through scheduled jobs and saves the collected data as plain Markdown files on disk — a format readable by both humans and models, without lock-in to a proprietary database. The first version ships with six connectors: Gmail, Notion, Git repositories, Twitter/X, Hacker News and web search, with Slack announced. Connectors are divided into deterministic (direct data feeds) and agentic (requiring intelligent source search).

Why local and Markdown matter

Storing data as Markdown on a local disk makes OpenWiki Brains a privacy-friendly alternative to cloud memory systems — data never leaves the user’s environment. This matters for organizations that want to give agents access to email and internal documents without sending that data to external services. It is distributed as an NPM package and integrated with the LangChain ecosystem (LangGraph, LangSmith).

Context

OpenWiki Brains arrives in a week when agent memory has become a central topic: Anthropic introduced Reflect for insight into one’s own habits, and Mistral added prompt governance to Studio. LangChain positions itself at the intersection of two trends — agentic memory and local, private processing — offering an open-source foundation instead of a closed product. The practical test will be how quickly the community adds new connectors, since the value of proactive memory grows with every new context source.