Use this page to pick the first resource to try. If you want the fastest path, start with RAM or disk. If you need real integrations, jump from the setup guide to the resource docs.The Access column separates filesystem access from API actions. read, write, and exec are MountMode values; exec allows commands to launch binaries from the mount and is typically used with Disk or RAM. actions means the resource also exposes mutating commands such as sending a message or creating an issue; run those commands on a WRITE mount. The Docs column shows where each resource is available, with one icon per supported runtime: Python, TypeScript (Node), and TypeScript (Browser). Click an icon to jump to that runtime’s docs.Sections below mirror the Setup sidebar so you can pick a category and follow the same path through credentials → resource docs.
Mirage drops into the major agent application frameworks. Each adapter exposes a Workspace as the agent’s filesystem and shell. Click an icon to jump to the integration docs.
Framework
Docs
Notes
OpenAI Agents SDK
MirageShell and MirageEditor plug into shellTool and applyPatchTool.
Vercel AI SDK
mirageTools() returns five typed tools for generateText / streamText.
LangChain (deepagents)
LangchainWorkspace backend for Python and Node deepagents.
Pi Coding Agent
Mirage extension for @mariozechner/pi-coding-agent. Node only.
Mastra
mirageTools() for Mastra Agent definitions. Node only.
Pydantic AI
For pydantic-ai and pydantic-deepagents.
CAMEL-AI
For CAMEL ChatAgent.
OpenHands
For the OpenHands agent SDK.
Claude Code (CLI)
Mount via FUSE; run claude against the mountpoint.
Home
Python
TS