Chroma

Chroma is an open-source search and retrieval database purpose-built for AI applications, designed to be the storage layer behind retrieval-augmented generation (RAG) pipelines, AI agents, and semantic search systems. The company was founded in 2022 by Anton Troynikov and Jeff Huber and is headquartered in San Francisco, raising an $18 million seed round in April 2023 led by Quiet Capital. Chroma became one of the most widely adopted open-source vector databases during the 2023 LLM boom, spreading primarily through a simple "pip install chromadb" developer experience and deep integration with popular frameworks like LangChain and LlamaIndex. Today it has more than 26,000 GitHub stars, is used in over 90,000 other open-source codebases, and is downloaded more than 11 million times a month, with production users including Capital One, Weights & Biases, UnitedHealthcare, Live Nation, Cisco, Intel, and Sony.

At the core of Chroma's product is a unified query interface spanning five search methods: dense vector search for semantic similarity, sparse vector search using BM25 and SPLADE for lexical precision, full-text and regex matching, and metadata filtering, all combinable through hybrid search for better retrieval quality than any single method alone. The same Apache 2.0-licensed codebase powers both the free, self-hosted single-node version (installable via pip, npm, or Docker, with in-memory or persistent storage) and Chroma Cloud, the company's fully managed, serverless offering that reached general availability in August 2025. Chroma Cloud is built on a Rust engine called Chroma Distributed, running on S3-style object storage with intelligent data tiering between memory, SSD cache, and cold object storage, and is SOC 2 Type 2 certified across AWS and GCP regions in the US and EU.

Newer capabilities extend Chroma well beyond a basic vector store: Forking lets teams duplicate a collection instantly for A/B testing or rollouts without re-indexing costs, Web Sync automatically crawls, chunks, and embeds entire websites (docs, changelogs, support pages) directly into a Chroma collection, and a Package Search MCP lets AI agents query thousands of open-source repositories directly. Technical specs for Chroma Cloud include support for up to 1 million collections per database, 5 million records per collection, 30 MB/s write throughput per collection, and warm query latencies around 20ms at p50, scaling to sub-2-second cold queries at p99.

Chroma's pricing is refreshingly simple for a usage-based infrastructure product: self-hosting is entirely free, and Chroma Cloud starts with $5 in free credits and no minimum spend, billed across four transparent usage dimensions (data written, data stored, data queried, and egress). This makes Chroma an excellent starting point for developers prototyping RAG systems and AI agents who want the simplest possible setup experience, while teams operating at very large scale or requiring the most mature multi-tenant SaaS isolation may still want to evaluate dedicated systems like Pinecone or Weaviate alongside it.

Self-hosting Chroma is completely free and open source under the Apache 2.0 license, installable via pip, npm, or Docker with no usage limits. Chroma Cloud, the managed serverless offering, gives new accounts $5 in free credits with no minimum commitment, then bills usage across four transparent dimensions: $2.50 per GiB written, $0.33 per GiB-month stored, $0.0075 per TiB queried, and $0.09 per GiB of egress. The Team plan includes $100 of usage credits that do not roll over month to month. Enterprise pricing is fully custom and adds features like SOC 2 Type 2 compliance guarantees, dedicated clusters, AWS PrivateLink connectivity, customer-managed encryption keys, and direct Slack support with custom SLAs; interested teams should contact Chroma's sales team directly. Credits generally do not expire outside of the non-rolling Team plan allocation.