AutoGen Studio

AutoGen Studio is a low-code / no-code developer GUI from Microsoft Research for rapidly prototyping, composing, and debugging multi-agent AI workflows built on the AutoGen framework. Shipped as the `autogenstudio` Python package and launched with `autogenstudio ui`, it serves a FastAPI + React (Gatsby) web app on localhost that exposes four primary interfaces — Team Builder, Playground, Gallery, and Deployment — backed by a SQLModel persistence layer (SQLite by default; any SQLAlchemy-compatible backend such as PostgreSQL, MySQL, MSSQL via `--database-uri`). The Team Builder offers drag-and-drop and JSON authoring of teams, agents, models, tools, and termination conditions fully aligned with AutoGen AgentChat's declarative component spec; the Playground streams live inter-agent messages and renders the control transition graph; the Gallery imports community components; and the Deployment view exports a team to Python, exposes it as an endpoint, and packages it for Docker. AutoGen Studio is built on AutoGen AgentChat / Core / Extensions and supports any OpenAI-compatible model endpoint (OpenAI, Azure OpenAI, Anthropic, local vLLM/Ollama, etc.) via declarative `model_client` configuration, plus MCP tool integration. Authentication is experimental (GitHub OAuth + JWT only). Microsoft explicitly positions AutoGen Studio as a research prototype — not production-ready — and encourages teams that need authn/z, multi-tenancy, sandboxing rigor, or hardened deployment to build directly on the AutoGen framework instead. Companion to the broader AutoGen multi-agent framework, distributed under the Microsoft microsoft/autogen monorepo (CC-BY-4.0 docs, MIT code).

Kin Score

This is the API Evangelist rating — a single, repeatable read computed from the artifacts on this page. Green fill is points earned; the red track is points possible, so every bar shows earned-versus-possible at a glance.

How we profile AutoGen Studio

Each block below is one kind of artifact we hold for AutoGen Studio. For each we say what it is and why it earns a place in the profile, then list every one we've indexed — capped at two rows, scroll within the panel for the rest.

Features 21

The notable capabilities this provider advertises, captured as structured features so they can be searched and compared instead of read one landing page at a time.

Notable capabilities this provider offers.

Installable via `pip install -U autogenstudio` (Python 3.10+); current PyPI release 0.4.2.2

Launched as a local web app with `autogenstudio ui --port 8081` (FastAPI backend + Gatsby/React frontend)

Configurable via `--host`, `--port`, `--appdir`, `--reload`, `--database-uri`, `--upgrade-database`, `--auth-config`

Team Builder visual canvas with drag-and-drop assembly of teams, agents, models, tools, and termination conditions, plus equivalent direct JSON editing

Component Library backed by AutoGen AgentChat's declarative component spec (teams, agents, models, tools, termination conditions)

Playground with live inter-agent message streaming, control transition graph visualization, UserProxyAgent sessions, and pause/stop run control

Gallery for discovering and importing community-created components and third-party integrations

Deployment view that exports a team to Python code, exposes it as a runnable endpoint, and supports containerized execution via Docker

Bring-your-own model — any OpenAI-compatible endpoint (OpenAI, Azure OpenAI, Anthropic, vLLM, Ollama, local models) via declarative `model_client` config; AutoGen Extensions provides first-party clients

Define agents in Python with AutoGen AgentChat, dump to JSON via `dump_component().model_dump_json()`, and import into Studio's JSON editor

MCP (Model Context Protocol) tool integration via `autogenstudio/mcp` and `/api/mcp` routes

SQLModel-based persistence (Pydantic + SQLAlchemy) — defaults to SQLite, supports PostgreSQL, MySQL, MSSQL, Oracle, and other SQLAlchemy dialects

Internal FastAPI surface (not a public API) under `/api/` with routes for teams, sessions, runs, gallery, mcp, settings, validation, and a `/api/ws` WebSocket for streaming

Experimental GitHub OAuth authentication with JWT (`--auth-config auth.yaml`); disabled by default, WebSockets require `?token=` query param when enabled

Default app directory `~/.autogenstudio/` for database and generated user files

Dev container shipped under `python/packages/autogen-studio/.devcontainer/` for source builds

{"Frontend stack" => "React + Gatsby (built with `yarn build`); backend served by `autogenstudio.web.serve`"}

Explicitly positioned by Microsoft as a research prototype — not production-ready; lacks production-grade authn/z, multi-tenancy, jailbreak hardening, and least-privilege key scoping

Companion to the broader AutoGen framework (AgentChat, Core, Extensions, .NET) under the microsoft/autogen monorepo (~58k GitHub stars)

Original research prototype (Oct 2023) by Dibia, Bansal, Fourney, Choudhury, Amershi, Awadallah, Wang; EMNLP 2024 System Demonstrations paper

{"License" => "MIT for code, CC-BY-4.0 for documentation"}

Scroll within the panel for all 21 ·

Resources

Every other property we hold for AutoGen Studio — documentation, portals, status pages, policies, and corporate surface — grouped by the job it does, following the integrator's arc from getting started to running in production.

← All providers · Data indexed from github.com/api-evangelist/autogen-studio · machine-readable index on apis.io