aakarim/OpenLore

What it solves

OpenLore solves the problem of fragmented and inconsistent knowledge across multiple AI agents. Instead of each agent having its own siloed context, OpenLore provides a centralized, governed, and version-controlled "knowledge surface" that multiple agents can read from and write to using standard tools.

How it works

OpenLore acts as a server that exposes a folder of Markdown files as a virtual filesystem. It supports multiple transport protocols, including SSH, SFTP, and the Model Context Protocol (MCP). Because it uses a virtualized Go-based shell rather than a native OS shell, it allows agents to interact with documentation using familiar commands like cat, grep, and find without risking security vulnerabilities like shell escapes. It enforces role-based access control, allowing administrators to define who can read, write, or publish to specific docsets.

Who it’s for

It is designed for teams of developers and AI agents that need a shared, durable memory. It is particularly useful for managing internal API documentation, runbooks, and project context that needs to be kept up-to-date across different coding assistants like Claude Code, Cursor, or automated CI pipelines.

Highlights

• Agent-native interaction: Agents use standard shell commands and composition patterns they already know.
• Multi-protocol support: Serves the same knowledge base over SSH, SFTP, MCP, and a web interface.
• Governed access: Provides granular, identity-scoped permissions (read-only, publish, or read-write) for different agents and users.
• Safe execution: Uses an in-memory Go interpreter to prevent arbitrary process execution or network access.
• No complex infrastructure: Requires no vector databases or ingestion pipelines; knowledge remains as simple, portable Markdown files.