The short answer
Most open-source AI tools in our catalog use permissive licenses, MIT or Apache 2.0, which allow commercial use with few conditions. A few use AGPL-3.0, which requires sharing changes if you offer a modified version as a network service, and some are source-available rather than open source, such as Arize Phoenix under the Elastic License 2.0. Several 'open core' projects add a separate license for enterprise features. Check the license before you build a product on top of a tool.
| License type | What it generally allows | Examples in our catalog |
|---|---|---|
| MIT | Use, modify and sell, keeping the copyright notice | llama.cpp, OpenCode, Kilo Code, OpenHands, Langflow |
| Apache 2.0 | Like MIT, plus an explicit patent grant and notice requirements | vLLM, Unsloth (core), goose, Qdrant, Milvus, Chroma, Mem0, Letta, Jan |
| AGPL-3.0 | Share source of modified versions, including when offered over a network | Firecrawl, Skyvern |
| Open core | Core under MIT, enterprise directories under a separate license | Langfuse, Activepieces, Tabby |
| Source-available | Code is public, but the license restricts some uses | Arize Phoenix (Elastic License 2.0), Open WebUI (branding terms), Weaviate (BSD-3-Clause and Weaviate License) |
Read this first
This guide summarizes licenses as stated in each project's repository on October 2, 2026. It is not legal advice; for a product decision, read the license text and ask a lawyer.
Permissive licenses: MIT and Apache 2.0
MIT and Apache 2.0 let you use, modify and redistribute the software, including in commercial products, as long as you keep the required notices. Apache 2.0 also includes an explicit patent license. Most of the open-source inference and data tools in our catalog use one of these two, including llama.cpp (MIT), vLLM (Apache 2.0), Qdrant (Apache 2.0) and OpenCode (MIT).
Copyleft for network services: AGPL-3.0
AGPL-3.0 requires that if you modify the software and let people use it over a network, you make your modified source available to them. Firecrawl and Skyvern use AGPL-3.0 for their open-source editions. Using them as separate services is common; building them into a closed product needs more care.
Open core: one repository, two licenses
Some projects license most of their code under MIT but keep enterprise features in separate directories under a commercial license. Langfuse, Activepieces and Tabby work this way. You can use the open parts freely; the enterprise parts need a paid license.
Source-available is not the same as open source
Arize Phoenix describes itself as open source but is licensed under the Elastic License 2.0, which is not an OSI-approved open-source license and restricts offering it as a managed service. Open WebUI uses BSD-3-Clause terms plus a rule that deployments above 50 users in 30 days must keep its branding. Weaviate's code is under BSD-3-Clause and its own Weaviate License.
These projects are often free to use internally. The restrictions matter if you resell the software, host it for others or rebrand it.
Before you build on an open-source AI tool
Read the LICENSE file in the repository, not the marketing page. Check whether the features you need sit in an enterprise directory. If you will host the tool for customers, check for AGPL or source-available terms. Record the license and version you adopted, since licenses can change in later releases.
Sources & method
Official documentation was checked on 2026-10-02. Selection advice is our assessment. Interfaces and requirements may change; check the linked source for the version you intend to use.
- MIT License (Open Source Initiative)
- Apache License 2.0 (Open Source Initiative)
- GNU AGPL v3 (Open Source Initiative)
- Elastic License 2.0
- Arize Phoenix on GitHub
- Open WebUI license
- Langfuse on GitHub
- Firecrawl on GitHub
How SOTA reviews projects · Suggest a correction · Download the evaluation checklist
