Self-hosted MCP file manager for AI-powered file access
filex, from BRF Tech, is a self-hosted file management platform that connects AI agents to local and cloud storage. It runs a built-in Model Context Protocol server and provides a web explorer that supports listing, reading, writing, and organizing files through an AI endpoint. Key elements include multi-storage drivers, a single Go binary deployment, folder-scoped API tokens, and client components for React and Vue. Developers and power users needing controlled AI access to files are the primary audience.
What tasks can you actually use it for?
The app combines a high-performance file explorer with a built-in Model Context Protocol server so AI agents can operate on repository content directly. Document workflows explicitly mentioned include reading codebases, generating or updating documentation, and reorganizing asset folders. These capabilities let LLMs perform file-centric automation where programmatic access to project files reduces manual navigation and copy-paste editing during development or content preparation.
How dependable are AI agent interactions for file operations?
AI interactions occur through an exposed MCP endpoint, and the product states compatibility with MCP-compliant clients such as Claude Desktop. The app provides confined API tokens that can be scoped to specific folders, which controls agent visibility. Real-time collaboration is supported via presence avatars and WebSocket updates, so agent-driven changes and human edits appear live in the interface during a session.
What storage backends and extensibility does it accept?
The app supports multiple storage drivers and allows mounting of several targets simultaneously. Supported types listed include:
- Local disk
- Amazon S3
- SFTP and FTPS
- NFS and WebDAV (SMB/NAS noted in documentation)
It also exposes a plugin API for custom drivers, enabling bespoke backends or enterprise storage integration when default drivers do not cover a use case.
Is it practical to host and embed into existing projects?
Deployment uses a single Go binary that runs on Windows, macOS, and Linux and is commonly deployed via Docker or the CLI, which supports self-hosting on private infrastructure. Embedding is supported through client-side components for Vue 3, React, and a framework-agnostic web component, and the web UI offers grid and list views, image previews, and live file updates for direct integration into other applications.
Practical choice for controlled AI file access with human oversight
filex is a practical choice for developers and power users who need managed AI access to project files hosted on their own infrastructure. Integration depends on an MCP-compatible client for agent-driven workflows, so teams should plan to review any AI-generated changes before deploying to production. The tool suits projects that require embedding a managed file interface while preserving administrative control and oversight.
