What are you using for a shared, agent-native knowledge base across your team?

I've been exploring different agent memory and context management tools over the past few months, including Mem0, Engram, Memori, and a few others. Most of my initial exploration was around persistent memory, retrieval, and retaining context across agent sessions, I have build few projects around those as well. But there's another related category I've been looking at: shared knowledge bases where both humans and AI agents can access and contribute to the same project or team context. Tools like ByteRover and Basic Memory are already approaching this in different ways, from selectively syncing project memories across agents and teammates to maintaining Markdown-based knowledge accessible through MCP. Recently came across another open-source project called OpenLore, and I've been exploring it myself. It takes a fairly straightforward approach: serve existing Markdown files as a shared knowledge base over SSH or MCP, with a web interface for humans. There is no vector database, ingestion pipeline or SDK. A few implementation details I found interesting: RBAC: Identities and roles determine which document collections (docsets) each person or agent can read, write, or publish to. Controlled contributions: Agents can be given permission to publish findings to an inbox without allowing them to modify existing documentation. File-based access: Agents can use familiar commands like grep, cat, and find, without needing a separate vector database or ingestion pipeline. What interests me about this category is that a team's knowledge is often spread across repositories, documentation, runbooks, architecture decisions, and individual agent sessions. There are different ways to make that knowledge available to agents. ByteRover focuses on syncing and recalling project context, Basic Memory combines Markdown with indexed and linked knowledge, and OpenLore takes more of a shared filesystem approach with scoped access. I'm not sure there's one best approach here. A shared Git repository with Markdown and AGENTS .md files might already be enough for some teams. But once multiple developers and agents are contributing to the same knowledge base, permissions, updates, and deciding what becomes trusted context seem worth thinking about.