Your local AI studio, in a single binary.
Chat and image generation with Gemini, OpenAI, Anthropic, and DeepSeek — with your own keys. The app, database, and API keys stay on your machine; prompts and content go only to the provider you select. No telemetry, no Node.
$ irm https://mangostudio.dev/install.ps1 | iex$ curl -fsSL https://mangostudio.dev/install.sh | bash$ bun add -g mangostudio$ npm i -g mangostudio$ scoop bucket add juliopolycarpo https://github.com/juliopolycarpo/scoop-bucket && scoop install mangostudio$ brew install juliopolycarpo/tap/mangostudio$ cargo install mangostudio$ docker run -p 3001:3001 -v mango-data:/data ghcr.io/juliopolycarpo/mangostudio
$ curl -fsSL https://mangostudio.dev/install.sh | bashPlanned channel — not available yetrequires Bun v1.3.14+ · macOS · Linux · Windows
MangoStudioonlineOne command boots the API (Elysia + SQLite) and the React UI — and opens the studio in your browser.
From zero to studio in 3 steps
Install
Pick the platform and installer that fit your machine. Every ready channel installs the same binary.
bun add -g mangostudioConfigure
Add your provider key in config.toml or via an environment variable.
export GEMINI_API_KEY=…Serve
Boot the API + UI and open the studio in your browser. Ready to chat.
mangostudio serveA complete studio, local-first
Multi-model chat
SSE streaming, extended reasoning, tools, and mid-conversation model switching.
Image generation
Generate images straight from chat with compatible models and a persisted local gallery.
Attachments
Upload files and images; MangoStudio hands them to the provider at the right turn.
Tools & agents
grep, glob, shell, read_file, and subagent delegation with its own lifecycle.
Keys & BYOK
Keep keys in the OS secret store, in config.toml, or in .env. You stay in control.
Local-first & SQLite
Chats and settings persist in local SQLite via Kysely. The app, database, and API keys stay on your machine; prompts and content go only to the provider you select.
Works with the providers you already use
Built in the open, with the community
Modular DDD architecture, end-to-end typed contracts, and docs to contribute. Issues and PRs are welcome.