Helicone

license

Apache-2.0(OSI)

source

public

core-gated

ungated(Apache-2.0 across the whole recursive tree with a single LICENSE, no ee/ or enterprise directory, no license-key or premium flag

Apache-2.0 throughout, self-hostable in full, with a hosted Helicone Cloud sold beside it. A separate hosted product is not a gated core, and the cloud's tiered pricing is not evidence on this dimension. The read behind that: the whole default branch contains exactly one LICENSE, the Apache-2.0 root file. There is no enterprise directory, no second license, no license-key module and no entitlement flag. The README's self-host path is the published docker-compose file covering all five services - web, worker, jawn, Supabase and ClickHouse, plus Minio. What does exist is plan gating, and it ships as published Apache-2.0 code: web/lib/features.ts declares the add-on and non-free feature lists, web/lib/freeTierLimits.ts sets the quotas, and web/hooks/useFreeTierLimit.ts reads them. Those enforce Helicone Cloud's billing plan against a hosted account, ship in the source, and are modifiable under Apache-2.0, so the thing that looks like a gate withholds nothing. One datum points the other way and is recorded rather than smoothed over: the pricing page lists SAML SSO on the Enterprise tier, and the repository contains no SAML, SSO, OAuth identity-provider, Okta or WorkOS implementation anywhere. That is an absence rather than a carve-out or a closed package, and the Enterprise tier's other named item, on-prem deployment, is support for a deployment the repository already ships for free. So it meets none of the tests for a gated core. It is the weakest point in this record, and a later reader who disagrees has the same facts in front of them.

• https://github.com/Helicone/helicone recorded 2026-06-04

  Apache-2.0 license; full source; hosted cloud paid tiers + self-host

• https://api.github.com/repos/Helicone/helicone/git/trees/main?recursive=1 recorded 2026-08-12

  The whole repository as one tree, `truncated: false`, so this is every path rather than a root listing. Searched for the gating mechanisms the rubric names: no path contains `/ee/`, none contains `enterprise` outside bifrost marketing pages and static images, there is exactly one `LICENSE` (the Apache-2.0 root file), and nothing matches `premium`, `entitle`, `saml`, `okta`, `workos` or `oauth`. The paths matching `tier` are packages/pricing/tiers.ts, web/lib/freeTierLimits.ts, web/hooks/useFreeTierLimit.ts, web/components/shared/FreeTierLimit*.tsx and two SQL migrations - all published.

• https://raw.githubusercontent.com/Helicone/helicone/main/README.md recorded 2026-08-12

  The vendor's own self-host instructions and architecture. "Helicone is simple to self-host and update. To get started locally, just use our docker-compose file", with a clone-and-`./helicone-compose.sh helicone up` recipe and no caveat about a reduced edition. It lists the services that make up the product - Web, Worker, Jawn, Supabase, ClickHouse, Minio - all of which are directories in this repository, and states "Helicone is licensed under the Apache v2.0 License". The one paid item named here is a Helm chart: "For Enterprise workloads, we also have a production-ready Helm chart available. To access, contact us at enterprise@helicone.ai" - packaging for a deployment the published docker-compose already performs.

• https://raw.githubusercontent.com/Helicone/helicone/main/web/lib/features.ts recorded 2026-08-12

  The plan gating, read rather than inferred from its file name. It declares `ADDON_FEATURES = ["evals", "experiments", "prompts"]` and `NON_FREE_FEATURES = ["sessions", "properties", "users", "datasets", "alerts"]`, with subfeature quotas resolved in web/lib/freeTierLimits.ts (10 prompt runs, 10 playground runs, 3 experiments and so on, each with an upgrade message). Every one of those features is implemented in this repository; the code decides what a Helicone Cloud account on the free plan may do, and it is itself Apache-2.0 and modifiable. Nothing is withheld from the published source.

• https://www.helicone.ai/pricing recorded 2026-08-12

  The paid tiers, listed for the record and against the dimension rather than for it. Pro $79/mo adds unlimited seats, alerts and reports, and HQL; Team $799/mo adds five organizations, SOC-2 and HIPAA compliance and a dedicated Slack channel; Enterprise adds a custom MSA, SAML SSO, on-prem deployment and bulk cloud discounts. These price a hosted account and a support relationship. The single item here the published source cannot build is SAML SSO, which is recorded in the note as the one datum arguing the other way.

6,058 GitHub stars on Helicone/helicone, which bands at 1K-10K stars and level 2. Stars cannot carry a product above level 3, because a star is not a use. No download, install, request or customer figure is published for this product, so stars remain the only honest signal and the level is directional.

• https://api.github.com/repos/Helicone/helicone recorded 2026-08-12

  stargazers_count = 6,058 for Helicone/helicone

A broad feature matrix - observability and agent tracing, prompt management versioned against production data, cost, latency and quality analytics, datasets and fine-tuning, and caching - with a strong LLM routing angle on top through an AI gateway spanning 100+ models with automatic fallbacks. That is level with Langfuse at 4, and it is held there because nothing here is a frontier-definer over the anchors.

• https://raw.githubusercontent.com/Helicone/helicone/main/README.md recorded 2026-08-13

  Feature grid - Observability, Agent Tracing, LLM Routing, Cost & Latency Tracking, Datasets & Fine-tuning, Automatic Fallbacks - under the heading "Helicone is an AI Gateway & LLM Observability Platform for AI Engineers". The AI Gateway offers "Access 100+ AI models with 1 API key through the OpenAI SDK", and Playground and Prompt Management are listed as their own items.

• https://github.com/Helicone/helicone recorded 2026-06-04

  feature list: observability/agent tracing, prompt mgmt, cost analytics, AI Gateway/routing, datasets/fine-tuning, fallbacks