Haystack

license

Apache-2.0(core)

commercial

Haystack-Enterprise

source

public

core-gated

gated(deepset names the free tier a Community Edition and sells Haystack Enterprise Starter with a private template/Helm repo plus early access to security features not yet in OSS)

Apache-2.0 self-hostable core with a commercial enterprise tier, and the gating is real. The repo itself is clean: a full untruncated recursive tree of deepset-ai/haystack at main carries one governing Apache-2.0 LICENSE and no enterprise or ee directory. What makes it gated is that the withheld material sits in a separate private repository rather than a subdirectory - Haystack Enterprise Starter ships "curated pipeline templates", a "Helm chart and guides", and "early access to enterprise-grade features ... like prompt injection countermeasures and other security-oriented features ahead of broader release". deepset's own tiering calls the free product "Haystack OSS (Community Edition)", and selling a commercial edition alongside is one of the things that counts as gating. This is deliberately distinguished from a hosted platform running the same published code: here it is code and features the published source does not contain at the time they are sold.

• https://github.com/deepset-ai/haystack recorded 2026-06-17

  Apache-2.0, ~25.6k stars, Haystack Enterprise tiers

• https://api.github.com/repos/deepset-ai/haystack/git/trees/main?recursive=1 recorded 2026-08-11

  Full untruncated recursive tree, 11,007 entries. Only license-related paths are the root `LICENSE`, `license-header.txt`, `licenserc.toml` and a `.github/workflows/license_compliance.yml`. No enterprise/, ee/, commercial/ or proprietary/ path anywhere.

• https://api.github.com/repos/deepset-ai/haystack recorded 2026-08-11

  Repo metadata: language Python, license spdx_id Apache-2.0, size 249,631 KB, 26,179 stars, default branch main.

• https://haystack.deepset.ai/blog/announcing-haystack-enterprise recorded 2026-08-11

  Lists what Haystack Enterprise Starter adds over OSS: 'Curated pipeline templates', 'Helm chart and guides for secure deployments', and 'Early access to enterprise-grade features - Take advantage of capabilities like prompt injection countermeasures and other security-oriented features ahead of broader release.' Also states 'Haystack ... will always remain an open source framework'.

• https://www.deepset.ai/blog/introducing-haystack-enterprise recorded 2026-08-11

  deepset's own three-tier framing: 'Haystack OSS (Community Edition): 100% open source, self-supported.' / 'Haystack Enterprise Starter: The power of Haystack with added enterprise pipeline and deployment templates, direct and private support' / 'Haystack Enterprise Platform (Cloud or on-prem)'. Pricing 'based on the size of your organization'.

933,408 PyPI downloads in the trailing 30 days for haystack-ai, the 2.x package, which bands at level 3 (100K-1M) on the software and model adoption scale. It sits just under the 1M boundary and could cross either way.

• https://pypistats.org/api/packages/haystack-ai/recent recorded 2026-08-12

  933,408 downloads in the trailing 30 days for haystack-ai

Most mature combined RAG + orchestration feature set among the peers. The repo publishes modular pipelines and agent workflows with explicit control over retrieval, routing, memory and generation, covering RAG, multimodal applications, semantic search, question answering and autonomous agents, and adds ready-made agents through Agent Pack and progressive skill discovery through SkillToolset.

• https://github.com/deepset-ai/haystack recorded 2026-08-13

  modular pipelines and agent workflows with explicit control over retrieval, routing, memory and generation; RAG, multimodal, semantic search, question answering and autonomous agents; Agent Pack ready-made agents and SkillToolset progressive skill discovery