Introducing Chroma Cloud

Chroma Cloud is a hosted, production-grade deployment of Chroma built from everything we learned shipping the single-node engine, re-architected into a distributed system for scale. Under the hood, it’s a cloud-native, distributed implementation of Chroma, but the API stays the same, so you can move from single-node to cloud without rewriting your app. Chroma Cloud runs across AWS and GCP and uses a distributed vector index to enable massive scale without provisioning or tuning infrastructure.

Distributed indexing#

Chroma Cloud uses an implementation of SPANN combined with SPFresh for our distributed vector index, enabling indexes to scale horizontally. Chroma Cloud’s strong consistency is driven by wal3, our object-storage backed write-ahead log.

In practice, this means you can:

• Ingest large collections without babysitting index operations.
• Query with low latency while your workload scales across a distributed infrastructure.
• Rely on a production-grade indexing path that’s tuned and operated for you (so you don’t have to become an indexing expert).

Dense, sparse, full-text, and metadata search#

Modern search and retrieval isn’t one-size-fits-all, Chroma Cloud supports a variety of index types:

• Dense vector search for semantic similarity over embeddings.
• BM25 and SPLADE for keyword-style search to capture exact term matches.
• Full-text search (FTS) and regex on documents when you need lexical matching over raw text.
• Metadata filtering with inverted indexes on metadata values to narrow results efficiently.

Collection forking#

Chroma Cloud Forking lets you duplicate a collection quickly without paying the time and cost of re-indexing. Forked collections share underlying data with the parent and only incur additional storage cost for incremental changes, keeping forks fast and cost-efficient.

This is ideal for:

• Versioned datasets (e.g., codebases by commit/tag)
• A/B testing retrieval settings and prompts against the “same” corpus
• Multi-tenant or per-customer variants without duplicating everything upfront

Dashboard#

Chroma Cloud includes a robust web based dashboard for day-to-day operations. Collaborate with your team, inspect and validate collection contents, and monitor usage and billing, all in one place.

• Collection configuration persists inside of Chroma’s dashboard, test and debug queries directly
• View performance metrics like p50, p90 and p99 query latencies as well as request and storage volumes

Insights and metrics#

In the dashboard, you can track key operational signals for your Cloud DBs and collections, including:

• Usage and activity so you can spot spikes, trends, and regressions.
• Storage and data growth to understand how collections are evolving as you ingest more data.
• Billing and spend visibility tied to usage so teams can stay on top of costs.

CLI#

The Chroma CLI is deeply integrated with Chroma Cloud so you can manage workflows from your terminal.

• Authenticate with chroma login, then use per-team profiles that persist API keys and tenant IDs locally
• Manage databases and explore data in collections with a built in collection browser
• Copy local data to and from Chroma Cloud: use chroma copy to copy one or more collections (or --all) from a local Chroma server into a Chroma Cloud DB.

Pricing#

Chroma Cloud is usage-based: you pay for what you use as you scale up (and scale down).

• No minimums, get started with production-grade infrastructure without upfront commitments.
• Predictable, transparent metering so teams can track and manage spend over time.

If you’re building anything from a prototype to a mission-critical production workload, Chroma Cloud is the fastest path to reliable, scalable retrieval, powered by the open-source Chroma engine, fully managed for you.