Chroma is an open-source embedding database built to give AI applications a way to store, search, and retrieve vector embeddings alongside metadata and text. Founded in 2022 by Jeff Huber and Anton Troynikov and based in San Francisco, Chroma shipped its first open-source release in February 2023 and has since become one of the most widely used vector databases in the retrieval-augmented generation ecosystem.
The project is dual-natured: a free, Apache 2.0-licensed open-source core that developers can run locally or self-host, and Chroma Cloud, a usage-based hosted service launched in August 2025 for teams that want managed, serverless vector search at scale.
Key Features
Chroma stores vector embeddings with associated metadata and supports similarity search, hybrid vector plus full-text search, and metadata filtering, making it suitable for retrieval-augmented generation, semantic search, and AI agent memory.
It can run embedded directly inside a Python or JavaScript application for local development, or be deployed as a self-hosted server for production, then migrated to Chroma Cloud's serverless infrastructure without changing the core API.
Pricing
The open-source, self-hosted version of Chroma is free with no usage limits. Chroma Cloud uses usage-based pricing: new accounts get $5 in free credits, storage is billed at about $0.02 per GB per month on object storage, and compute scales with query and write volume.
A Team-tier plan runs around $250 per month with $100 of included usage credits, and Enterprise plans offer dedicated infrastructure with negotiated, configurable billing for large-scale deployments.
Key Features
Vector similarity search — Stores and queries high-dimensional vector embeddings to find semantically similar items for RAG, search, and recommendation use cases.
Hybrid search — Combines vector similarity search with full-text keyword search in a single query for more precise retrieval.
Metadata filtering — Filters query results using structured metadata alongside vector similarity, narrowing results by attributes like date or category.
Local-first development — Runs embedded inside a Python or JavaScript/TypeScript application for fast local prototyping before deploying to production.
Self-hosted deployment — Can be self-hosted as a standalone server under the free, open-source Apache 2.0 license with no vendor lock-in.
Chroma Cloud (serverless) — A managed, usage-based hosted platform built on object storage that scales compute automatically with query and write volume.
Framework integrations — Works with common LLM application frameworks such as LangChain and LlamaIndex, and standard embedding model providers.
Pros & Cons
Pros
Genuinely free and open source (Apache 2.0) for self-hosted use, with no artificial feature gating
Simple local-first developer experience that scales to a managed cloud service without an API rewrite
Object-storage-based Chroma Cloud pricing is notably cheaper for storage than memory-based vector databases
Large existing adoption (tens of thousands of open-source projects) means broad community support and integrations
Cons
Chroma Cloud is usage-based, so costs can be harder to predict at very large scale than flat-rate plans
As a relatively young company (founded 2022), the hosted product has a shorter production track record than some incumbents
Advanced enterprise features (dedicated infrastructure, custom SLAs) require negotiated Enterprise pricing rather than self-serve signup
Small team size relative to larger, VC-scaled vector database competitors
Pricing
Open Source (Self-Hosted) Free N/A
Chroma Cloud Usage-based ($5 free credits to start) Pay-as-you-go
Team ~$250/month (includes $100 usage credits) Monthly
Enterprise Custom Custom
Frequently Asked Questions
Is Chroma free to use?
Yes. The open-source, self-hosted version of Chroma is free under the Apache 2.0 license. The hosted Chroma Cloud service is usage-based, starting with $5 in free credits for new accounts.
What is Chroma used for?
Chroma is primarily used to power retrieval-augmented generation (RAG) for LLM applications, semantic search, AI agent memory, and recommendation systems that rely on comparing vector embeddings.
Who founded Chroma?
Chroma was founded in 2022 by Jeff Huber, who serves as CEO, and Anton Troynikov. The company is based in San Francisco, California.
How is Chroma Cloud priced?
Chroma Cloud bills usage-based: storage costs about $0.02 per GB per month on object storage, and compute scales with query and write volume. A Team plan is priced around $250/month with $100 of included credits.
Can I self-host Chroma instead of using Chroma Cloud?
Yes. Chroma's open-source core can be self-hosted as a standalone server, run embedded in an application, or migrated later to the managed Chroma Cloud service.