
ChromaDB: Free and Open Source Tool
Open-source embedding database for building AI applications with semantic search.
Premium: $79/year for unlimited deals
Already have an account? Log in
Deal Highlights
Our ChromaDB listing: Free & Open Source
Offer: Free & Open Source. Our listing sets no requirements and there is no promo code. There is also no SaaSOffers tracking link for this tool, so step 1 of claiming it is simply going to Chroma's own site.
We checked these plans and prices against Chroma's official pricing, documentation and security pages in November 2026.
Chroma is an open-source database for embeddings, the number lists that AI models produce from text, images or other data. You store each embedding next to its document and metadata, then ask for the closest matches to a query. That is the retrieval step behind RAG chatbots, semantic search and recommendation features.
The software is free under the Apache 2.0 license, and Chroma's hosted service, Chroma Cloud, has a $0 Starter plan with usage billed on top. Chroma does not publish a startup or nonprofit discount program on its pricing page, so this listing adds no extra deal. What you get here is the public free route, explained with numbers.
Apache 2.0, 29,653 stars, and no telemetry since 1.5.4
Chroma's GitHub repository (chroma-core/chroma) reports the license as Apache 2.0 and showed 29,653 stars on November 8, 2026. GitHub's "latest release" entry showed version 1.5.9, published on May 5, 2026.
Chroma's docs say it "will always remain open-source (Apache 2.0)" and describe two engine modes:
- Single-node local Chroma, which is what
pip install chromadbgives you. - Distributed Chroma, which is the system Chroma Cloud runs.
The two use different storage subsystems today. The docs warn that local Chroma "may temporarily lack some features or behaviors available in the distributed system," and say most engineering effort now goes to the distributed system and the cloud. Reaching parity is a stated goal, not a finished fact. If you prototype locally and later move to Cloud, test your queries again on Cloud.
As of version 1.5.4, Chroma no longer collects product telemetry. You can wire up OpenTelemetry on your own instances, and the data goes to you, not to Chroma.
Four ways to run it, from a Python process to a hosted cluster
Chroma's getting-started page lists four modes:
| Mode | How you start it | Data survives a restart? |
|---|---|---|
| In-memory | chromadb.Client() in Python | No, data is lost when the program ends |
| Persistent | A persistent client pointed at a folder | Yes |
| Client-server | chroma run --path ./getting-started, or the chromadb/chroma Docker image | Yes, on the server's path |
| Chroma Cloud | CloudClient in Python or TypeScript | Yes, hosted |
Install commands from the docs: pip install chromadb, poetry add chromadb or uv pip install chromadb for Python. For TypeScript, add chromadb and @chroma-core/default-embed with npm, pnpm, bun or yarn. For Rust, cargo add chroma. The Docker image is chromadb/chroma.
Clients exist for Python, TypeScript and JavaScript (ESM and CJS), and Rust. The Rust client does not embed text for you, so you supply your own vectors there. In Python and TypeScript, Chroma can embed text when you add or query documents, using an embedding function interface that supports several providers. The docs page we read does not name a default model, so check which one your install uses before you depend on it.
Collections hold embeddings, documents and metadata. IDs must be unique strings, and a text query returns the 10 nearest documents by default.
Chroma Cloud: $0 plus usage, $250 plus usage, or custom
Chroma Cloud has three plans on its pricing page:
| Plan | Monthly price | Credits | Databases | Team members | Listed extras |
|---|---|---|---|---|---|
| Starter | $0 + usage | $5 in free credits | 10 | 10 | Community Slack |
| Team | $250 + usage | $100 in credits | 100 | 30 | Slack support, SOC II, volume-based discounts |
| Enterprise | Custom | None listed | Unlimited | Unlimited | Dedicated support, single-tenant clusters, SLAs |
The page shows no yearly rate. Starter and Team bill by credit card, and mid-cycle plan changes are prorated to the calendar month.
How much free credit Starter gives
Chroma's pages disagree slightly. The pricing page heading reads "per month + $5 in free credits." The getting-started and billing docs say new users receive $5 in credits. We could not confirm from Chroma's text whether the $5 repeats every month or is a one-time grant. The pricing FAQ says that "in most cases" unused credits do not expire, and that the $100 included in Team does not roll over. Ask Chroma support before you plan around $5 a month.
What each unit of usage costs
| Meter | Rate |
|---|---|
| Writes | $2.50 per logical GiB |
| Storage | $0.33 per GiB per month, prorated hourly |
| Queries | $0.0075 per TiB queried |
| Data returned | $0.09 per GiB |
| Forks | $0.03 per fork request |
| Sync processing | $0.04 per GiB |
| Document page extraction | $0.01 per page |
| Web Sync | $0.01 per page scraped |
Writes are billed on the raw, uncompressed size of what you send through add, update or upsert. Each write is billed once. Background compaction and reindexing are not billed. Forks copy on write, so after a fork you pay only for the new storage you write. Billing is per account, and usage across all your collections and tenants is added together.
A query is counted in units. A plain vector similarity query is 1 unit. Each metadata or full-text predicate adds 1 unit. Full-text and regex filters count as the number of characters in the search string minus 2. Chroma's example: 100,000 units against 10 GiB costs about $7.50.
Chroma's pricing page also has a calculator example: 1.0M documents written (13 GiB) for $34, 6.0M documents stored (80 GiB) for $27, and 10.0M queries at 4 queries per second for $19, which totals $79 a month. We could reproduce the first two lines with the published rates. We could not reproduce the $19 for queries with the unit rule above, so treat that line as Chroma's estimate and run your own numbers.
If a usage limit is hit, either one you set or one Chroma sets, service is paused until the limit is raised.
What the free route leaves out
On Starter, the pricing page lists community Slack as the only support channel. It lists no SOC II on Starter (that appears under Team), no volume discounts, no SLA and no single-tenant clusters. Starter allows 10 databases and 10 team members.
Self-hosting costs your own time. You handle:
- Backups of the persistent folder or server volume
- Upgrades, including the engine differences noted above
- Capacity and uptime on your own infrastructure
Chroma Sync (imports from S3, GitHub, the web or file upload), the Chroma CLI and the Search Agent are available only in the us-east-1 region.
Limits that matter before you design a schema
Chroma Cloud's quotas page lists these limits, and says most can be raised on request. It describes no per-plan differences, so ask support whether your plan changes them.
| Limit | Value |
|---|---|
| Embedding dimensions | 4,096 |
| Document size | 16,384 bytes |
| ID size | 128 bytes |
| Record metadata value | 8,182 bytes |
| Metadata keys per record | 32 |
| Where predicates per query | 8 |
| Results per query | 300 |
| Full-text or regex input | 256 characters |
| Concurrent reads per collection | 10 |
| Concurrent writes per collection | 10 |
| Records per write | 300 |
| Records per collection | 5,000,000 |
| Collections | 1,000,000 |
The 16,384-byte document cap means you store chunks, not whole files, so a 50 KB page needs at least 4. A write batch is 300 records, so loading 1,000,000 records takes at least 3,334 write calls.
Two worked examples with the published rates
Both use 1,536-dimension float32 embeddings, which take 1,536 x 4 = 6,144 bytes, plus a 1,024-byte document and 64 bytes of metadata. That is 7,232 bytes a record. We assume query results return 10 documents of 1,024 bytes each. We are assuming Chroma bills these sizes as logical bytes written, which is how the docs describe it.
200,000 records, 100,000 queries a month
- Size: 200,000 x 7,232 = 1,446,400,000 bytes, or 1.35 GiB
- One-time write: 1.35 x $2.50 = $3.37
- Storage: 1.35 x $0.33 = $0.44 a month
- Queries: 100,000 units x (1.35 / 1,024 TiB) x $0.0075 = $0.99
- Data returned: 100,000 x 10 KB, about 0.95 GiB, x $0.09 = $0.09
First month: about $4.89. That is inside a $5 credit if the credit applies to that month, and later months cost about $1.52 without any new writes.
2,000,000 records, 100,000 queries a month
- Size: 14,464,000,000 bytes, or 13.47 GiB
- One-time write: 13.47 x $2.50 = $33.68
- Storage: 13.47 x $0.33 = $4.45 a month
- Queries: 100,000 units x (13.47 / 1,024 TiB) x $0.0075 = $9.87
- Data returned: $0.09
First month: about $48.09. Steady months with no new writes: about $14.41. The query line scales with data size, not just query count. Team at $250 a month would cost far more than this usage, so Starter is the right plan here. Even with Team's $100 credit counted against usage, you would pay $250 for roughly $14 of use.
Team makes sense for the features, not the usage: $250 with $100 in credit is $150 net for Slack support, SOC II and 100 databases, if you use the full credit.
Regions, security and what Chroma does not say
Chroma Cloud has two regions: AWS us-east-1 (N. Virginia, endpoint api.trychroma.com) and GCP europe-west1 (Belgium, endpoint europe-west1.gcp.trychroma.com). Each database stays in the region where it was created. To move US data to the EU, you create a new database in europe-west1 and reindex.
The default is a fully managed multi-tenant cluster. Single-tenant deployments managed by Chroma are available on request, and the Enterprise plan lists single-tenant clusters.
Chroma's security page says Chroma Cloud completed a SOC 2 Type II examination. It states data is encrypted in transit and at rest. The page does not state HIPAA support or bring-your-own-cloud. The cloud docs say data is stored and processed within the region you select. The pricing page links a DPA, but we did not read its text. If you sell to regulated customers, ask in writing.
ChromaDB next to Weaviate, Qdrant and Pinecone
| Chroma | Weaviate | Qdrant | Pinecone | |
|---|---|---|---|---|
| Open-source engine | Yes, Apache 2.0 | Yes, BSD-3-Clause outside the wl folder | Yes, Apache 2.0 | No |
| Free hosted tier | Starter: $0 + usage, $5 in credits | 100,000 objects, 1 GB memory, 10 GB disk, 1 collection | 1 node, 0.5 vCPU, 1 GB RAM, 4 GB disk | Starter: up to 2 GB storage, 2M write units, 1M read units a month |
| Paid step | Team $250/month + usage | Flex, $45 a month minimum | Usage based, hourly | Builder $20 a month; Standard $50 a month minimum |
| Billing unit | Written GiB, stored GiB, TiB queried, GiB returned | Vector dimensions, storage, backup | vCPU, memory, disk reserved by the hour | Write units, read units, storage |
| Startup offer | None published | Discounted rate through a program or VC | 20% off the first year, by application | Startup program, credits not quantified |
Figures for the other three come from their own official pricing pages, checked in November 2026. Chroma's entry plan has no fixed minimum, while Weaviate Flex starts at $45 a month and Pinecone Builder at $20. Qdrant's free cluster is suspended after 1 week unused, and the Chroma pages we read mention no idle suspension. Team at $250 is the highest fixed monthly price in this table, though the others bill by usage, so the real comparison depends on your load.
Who it's for
- A founder building a first RAG prototype who wants to be querying in minutes with
pip install chromadb - Teams that want an open-source engine under Apache 2.0, with no license fee and no telemetry
- Teams that prefer pay-per-use hosting with a $0 entry plan to a reserved-cluster model
Who should skip it
- Anyone who needs HIPAA or bring-your-own-cloud today. Chroma's security page does not state them.
- Teams that need a support contract or SLA without paying. Only Team and Enterprise list support beyond community Slack, and SLAs appear only on Enterprise.
- Workloads that need more than 5,000,000 records in one collection without a raise request.
Setup tips
- Start with
pip install chromadband a persistent client, so restarts do not wipe your test data. - Chunk documents under 16,384 bytes before you embed them.
- Keep metadata under 32 keys per record and 8 predicates per query.
- Batch writes in groups of 300 records and run up to 10 in parallel per collection.
- Record your embedding model and dimension (limit 4,096) at collection creation.
- Set a customer usage limit in the Cloud dashboard, since hitting a limit pauses service.
- Re-test queries on Cloud after local development.
- Create your Cloud database in europe-west1 from day one if your users are in the EU. Moving later means reindexing.
How to claim the deal
- Go to trychroma.com. Our listing has no SaaSOffers tracking link for Chroma.
- For self-hosting, run
pip install chromadbor pull thechromadb/chromaDocker image. No account is needed. - For Chroma Cloud, create an account and choose the Starter plan, which is $0 plus usage.
- Check the credit shown in your billing page, since Chroma's pages describe the $5 credit two ways.
- Add a card only when you want usage beyond your credit. Starter bills by card.
- If you outgrow Starter, compare Team ($250 plus usage, with $100 in credits that do not roll over) against your real monthly usage first.
Who Is This Deal For?
Early-Stage Startups
Seed and pre-seed companies looking to move fast without overspending on tools.
Growing SaaS Teams
Series A+ companies scaling their stack and optimizing software costs.
Solo Founders
Indie hackers and bootstrapped founders who need enterprise tools at startup prices.
Get Free & Open Source off ChromaDB
Premium deal. Upgrade once, unlock everything.
Frequently Asked Questions
Everything you need to know about this startup deal.
Get the weekly deals digest
New verified startup deals every week. No spam, ever. Unsubscribe anytime.
Related Offers
Consensus
Used by 1,544 members
Free Plan
AI-powered research platform that finds, summarizes, and synthesizes scientific papers.
View offerLaxis
Used by 2,176 members
Free Plan
AI meeting assistant that records, transcribes, and generates actionable meeting notes.
View offerReclaim.ai
Used by 1,063 members
Free Plan
AI-powered scheduling and time management tool that automatically finds time for your priorities.
View offerReplicate
Used by 1,772 members
$500 in credits
Run open-source ML models in the cloud, deploy Llama, Stable Diffusion, and custom models via API without GPU management.
View offerZoomInfo
Used by 2,014 members
Free Trial
B2B data platform for sales and marketing intelligence with contact databases.
View offerDeal Summary
Looking for more startup deals?
Browse all offers