What it is
Chroma is an open-source database for storing embeddings and searching by meaning, widely used to prototype AI apps in Python and JavaScript. It runs on your own machine with a single install and no server, and when you grow you can move to Chroma Cloud, its serverless version. Beyond vector search, it offers full-text search and metadata filters.
How to use it
- Install the library in Python or JavaScript and store your first documents in minutes.
- Let it compute embeddings with its default function or the provider you choose.
- Filter by metadata and combine vector and full-text search.
- Move to Chroma Cloud when you need to scale without managing servers.
What it's good for
- A developer prototyping a chatbot over company PDFs in an afternoon.
- A team using CrewAI, which relies on Chroma for agent memory.
- An edtech startup finding similar exercises in its question bank.
Pros
- Open source (Apache 2.0) and free to self-host
- Runs locally with no server, great for prototypes and tests
- Works with LangChain, LlamaIndex and CrewAI
- Serverless cloud billed by usage, with no clusters to size
Cons
- Local mode isn't meant for large production workloads
- Fewer enterprise features than more established alternatives
- Docs are English-only
