Mirage - A Unified Virtual File System for AI Agents

Mirage is a unified virtual file system for AI agents. It mounts backends like S3, Google Drive, Slack, Gmail, and Redis into a single file system tree, allowing agents to access them through the same interface.

Agents handle resources using only bash tools, without needing to separately learn service-specific SDKs or MCP. It operates as a simulation environment, making multiple services appear like a local disk, and pipelines between services can be composed as file manipulation flows.

Key features are as follows.

• Multi-resource mounting: Supports RAM, Disk, Redis, S3/R2/OCI/Supabase/GCS, Gmail/GDrive/GDocs/GSheets/GSlides, GitHub/Linear/Notion/Trello, Slack/Discord/Telegram/Email, MongoDB, SSH, and more.
• Portable workspaces: Environments can be cloned, snapshotted, and versioned, so they can be moved to another machine and continue running without reconfiguration.
• App embedding: With Python and TypeScript SDKs, the virtual file system can be embedded directly inside FastAPI, Express, or browser apps.
• Framework compatibility: Integrates with OpenAI Agents SDK, Vercel AI SDK, LangChain, Pydantic AI, CAMEL, and OpenHands.
• Lightweight CLI + daemon: Can be attached to coding agents like Claude Code and Codex, letting them access mounted resources through familiar bash.
• Command extension and caching: New commands can be registered with ws.command(), and command overrides per resource and file type are supported. It uses a two-tier structure of index cache and file cache, with a choice of RAM or Redis backend.
• Pluggable backends: The default RAM configuration has a 512MB file cache and a 10-minute index TTL, while Redis shares the cache across workers, processes, and machines, and persists after restarts.

It has been released under the Apache-2.0 license.