Awesome AI AgentsAI Developer Toolkit

strukto-ai/mirage

⭐ 3627 TypeScript repository created 2026-05-06

Mirage is a unified virtual file system for AI agents. It mounts services and data sources such as S3, Google Drive, Slack, Gmail and Redis side by side under a single root, so any model that already knows bash can read, grep and pipe across every backend without learning new tool vocabulary. A workspace is declared by mapping mount points to resource objects, after which shell-style commands are executed against it: grepping a Slack channel, copying a file from S3 to local storage, or piping the output of find into wc. Commands are extensible, so new commands can be registered and existing ones overridden per resource and per file type; the project deliberately ships no file-type renderers, meaning a format such as Parquet renders however the user registers it. Around fifty built-in backends are available, spanning RAM and disk, Redis, object stores including S3, R2, OCI, Supabase and GCS, Google Workspace services, project trackers such as GitHub, Linear, Notion and Trello, messaging via Slack, Discord and email, databases including MongoDB, GridFS, Postgres, LanceDB and Qdrant, and SSH. Workspaces are portable: they can be cloned, snapshotted and versioned, so agent runs move between machines without restarting or reconfiguring. The SDKs are embeddable and run in-process inside FastAPI, Express, browser applications or any async runtime, with no separate process required. Python and TypeScript packages are published, the Python package also providing a CLI, and browser or edge builds are available. Agent integrations are offered for the OpenAI Agents SDK, Vercel AI SDK, LangChain, Pydantic AI, CAMEL and OpenHands through the SDKs, while coding agents connect through native adapters, installable plugins, MCP or FUSE mounts. Python 3.11 or newer and Node.js 20 or newer are required, on macOS or Linux.

https://github.com/strukto-ai/mirage

virtual-filesystemagent-toolsdata-integrationsdkmcp

Also in AI Developer Toolkit

Leonxlnx/taste-skill

Taste Skill provides portable AI Agent Skills designed to enhance the quality of AI-generated frontends and image-based design references, aiming to eliminate generic UI outputs.

unslothai/unsloth

Unsloth is a toolkit for fine-tuning and reinforcement learning of large language models, enabling faster training with significantly reduced VRAM usage across various advanced models and tasks.

reworkd/AgentGPT

AgentGPT is an open-source platform that allows users to assemble, configure, and deploy autonomous AI agents in a web browser to pursue custom goals by generating and executing tasks independently.

Gitlawb/openclaude

OpenClaude is an open-source command-line interface (CLI) coding-agent that supports various cloud and local model providers for unified AI-driven development workflows.

virgiliojr94/book-to-skill

A command-line tool that converts technical books, document folders or globs into structured agent skills in the open SKILL.md format, so coding agents load only the chapter relevant to a question.

google/adk-python

An open-source Python toolkit by Google for building, evaluating, and deploying sophisticated AI agents with flexibility, modularity, and integration within the Google ecosystem.

can1357/oh-my-pi

Oh My Pi is an AI coding agent for the terminal featuring hash-anchored edits, an optimized tool harness, LSP integration, and support for Python, browsers, and subagents.

millionco/react-doctor

React Doctor is an intelligent agent designed to diagnose and fix issues within React codebases, leveraging AI to improve code quality and developer efficiency.