Apache-2.0 LLM observability and AI gateway with proxy or async logging; Hobby is free at 10k requests a month with 7-day retention.
Teams that want the whole stack under a single OSI licence - the repository LICENSE is Apache License 2.0 - and are willing to run it themselves with Docker Compose across Web, Worker, Jawn, Supabase, ClickHouse and MinIO. The proxy integration suits teams who want request logging without rewriting application code, since traffic routes through Helicone by changing a base URL, and the AI Gateway covers a long list of providers from one endpoint. Sensible where cost and latency per provider are the questions being asked, rather than deep multi-step agent evaluation.
Retention is the deal-breaker to check first: 7 days on the free Hobby plan, 1 month on Pro at $79/month, 3 months on Team at $799/month, and configurable retention only on Enterprise. That is a steep ladder if you need quarter-over-quarter history. Volume forecasting is also hard, because no per-request or per-GB overage rate is published - the page says only that usage-based pricing applies past 10K requests and 1 GB, and shows a calculator figure of $0.97 for 10K requests and 0.30 GB without the underlying rate. Rate limits bite early too: Hobby allows 10 logs/min and 10 API calls/min, with 1 seat and 1 organisation. On portability, Helicone does not document a generic OpenTelemetry OTLP ingestion endpoint; the async path runs through Helicone's own @helicone/async and helicone-async SDKs wrapping OpenLLMetry, so switching backends later is more work than with an OTLP-native tool. SAML SSO, custom MSA, supported on-prem deployment and configurable retention are Enterprise-only.
Pros
Cons
Yes by licence. The repository LICENSE file is the Apache License, Version 2.0, and the README repeats that Helicone is licensed under Apache v2.0 with no carve-out mentioned for any directory. Self-hosting is documented with Docker Compose running Web, Worker, Jawn, Supabase, ClickHouse and MinIO. A production Helm chart for enterprise workloads is available by contacting Helicone, and manual deployment is explicitly not recommended.
7 days on the free Hobby plan, 1 month on Pro at $79 a month, and 3 months on Team at $799 a month. Only Enterprise keeps data indefinitely with configurable retention policies. If you need more than a quarter of history on a hosted plan you are into an Enterprise conversation, which is one of the clearer arguments for running the Apache-2.0 build on your own ClickHouse.
Pro is $79 a month and Team is $799 a month, each including 10,000 requests and 1 GB of storage before usage-based pricing applies. The per-request and per-GB rates behind that are not published; the pricing page only shows a calculator result of $0.97 for 10,000 requests and 0.30 GB. Both paid plans include a 7-day free trial, and seats are unlimited.
Not as a documented OTLP endpoint. The docs offer an async logging route described as using OpenLLMetry, but the setup shown uses Helicone's own SDKs - @helicone/async for Node and helicone-async for Python - initialised with HeliconeAsyncLogger and an API key, and no OTLP URL or exporter configuration is published. If OpenTelemetry portability is a requirement, Langfuse, Braintrust and Phoenix document it explicitly and Helicone does not.
Facts sourced from the vendor’s own site, last verified 2026-10-10.
The Weekly Source
One email a week. New listings, repriced tools, and the ones we would skip - including this one if its pricing changes.
