Mirage
A virtual terminal for AI agents that mounts services such as S3, Google Drive, Slack and Gmail as one filesystem, provides virtualized command-line tools, and gates what an agent can run or see through a policy engine.
Mirage is a virtual terminal for AI agents, combining a virtual filesystem, virtualized command-line tools, and routed dynamic runtimes into one interface. A virtual filesystem mounts backends such as S3, Google Drive, Slack, Gmail and Redis side by side under one root, so an agent can reach all of them with familiar Unix tools such as ls, grep, find and jq instead of separate SDKs or MCP servers for each.
Virtual command-line tools such as git, slack and its own ntn command are answered by Mirage itself, so an agent can drive a service with nothing installed, and one tool can be virtualized under multiple names with separate credentials; Python, JavaScript or other commands are routed to a configured runtime that can run in-process, sandboxed or remotely. Access is controlled with allow, ask and deny rules for commands and CLIs, and hide and show rules for files, so a hidden path is entirely absent from what the agent can see, all enforced by a scriptable policy engine that also gates every filesystem operation and session write. External changes, such as a new Slack reply, surface as filesystem change events instead of requiring the agent to rescan.
It is written in TypeScript and licensed under Apache-2.0.