Getting Started¶
Prerequisites¶
- Python 3.11+
- Node.js 18+
- Docker Desktop — download here (used to run MongoDB)
Mozaiks installs Tree-sitter parser packages with the core framework checkout so Studio can build source-backed App Intelligence for generated and existing apps. Mozaiks is not published on PyPI yet, so install it from this repo instead of pip install mozaiks. No separate parser setup is required.
1. Install¶
2. Start MongoDB¶
Mozaiks requires a database to store your apps, build history, and workspace state. The easiest way to start one is via Docker:
Don't want Docker?
Use MongoDB Atlas (free tier, cloud-hosted, nothing to install). Set MONGO_URI to your Atlas connection string in step 3.
3. Set environment variables¶
Free LLM key — Google Gemini
The default provider is Google Gemini, which has a free tier (no credit card required). Get your key at aistudio.google.com → Get API key.
Other providers
Using Atlas? Replace MONGO_URI with your Atlas connection string.
Prefer OpenAI? Set OPENAI_API_KEY=sk-... and LLM_PRIMARY_API_TYPE=openai instead.
Using Anthropic? Set ANTHROPIC_API_KEY=sk-ant-... and LLM_PRIMARY_API_TYPE=anthropic.
4. Create your workspace and open Studio¶
Replace my-workspace with whatever you want to name your app folder:
This scaffolds the workspace, starts the backend and frontend, and opens Studio in your browser at http://localhost:3000.
5. Build your first app¶
- Click Create App
- Describe what you want to build in the chat
- Follow the workflow — the AI guides you through the build steps
- Review and promote the generated app when it's ready
In-progress builds stay in Apps so you can always pick up where you left off.
This first build is your app's Genesis Build; after that, every change you ask for is a Refinement Run against the app you already have. See Genesis Builds and Refinement Runs.
Code Context Infrastructure¶
Mozaiks indexes app source into a SourceContextBundle, AppContextGraph, and AppIntelligenceSnapshot after generated artifacts, App Intelligence indexing, or existing-app discovery. This is automatic and uses Tree-sitter-backed parsing where supported.
For local quickstarts, graph snapshots are stored with Mozaiks artifacts. For team or production deployments with large repositories, plan to run FalkorDB as the graph-query mirror for faster multi-hop queries and Studio visualization. FalkorDB mirrors canonical AppContext and App Intelligence artifacts; it does not replace them.
Coming back to an existing workspace
Troubleshooting¶
MongoDB connection error
Make sure Docker is running and the MongoDB container is up:
If you used Atlas, double-check the connection string is correct and the cluster is reachable.
mozaiks is not recognized
Use python -m mozaiks instead. The mozaiks shortcut requires your Python scripts directory to be on PATH, which some systems don't configure automatically.
Builds fail or hang
Make sure your LLM API key is set in the current shell session. The default provider is Google Gemini — set GEMINI_API_KEY. If you switched to OpenAI or Anthropic, set OPENAI_API_KEY or ANTHROPIC_API_KEY respectively. You can open Studio without a key but builds will not run.
What's Next¶
-
Key Concepts
Workspaces, apps, modules, workflows — the full mental model in one page.
-
CLI Reference
Every
mozaikscommand in one place. -
Use Studio
Learn the workspace and app dashboard surfaces.
-
Guides
Add modules, pages, workflows, integrations, and more.
Contributing to Mozaiks or setting up from a repo checkout? See Local Setup.