Helicone

Helicone is two products under one brand. The core is an LLM observability platform: one line of code (swap the base URL) logs every request going through OpenAI, Anthropic, and 100+ other providers, then gives you traces, per-user cost breakdowns, a query language over your logs (HQL), prompt versioning, and a playground. Layered on top is the AI Gateway: a single OpenAI-compatible endpoint (ai-gateway.helicone.ai) that lets you call 100+ models without holding separate provider keys, with automatic fallbacks and load-balancing across providers. Both are usable from the same account; the gateway is the newer, higher-growth part of the pitch.

The free tier is real, not a teaser: 10,000 requests/month with full observability and no credit card, confirmed on the current pricing page (https://helicone.ai/pricing). The gateway's advertised 0% markup on provider tokens — you pay exactly what OpenAI/Anthropic charge — is a genuine differentiator against OpenRouter's 5.5% markup, per Helicone's own comparison table on their gateway docs (https://docs.helicone.ai/gateway/overview); we haven't independently re-priced every model to confirm the OpenRouter figure, so treat the 5.5% specifically as Helicone's claim about a competitor. The core platform repository is healthy by any open-source measure: 6,176 stars, 674 forks, Apache-2.0, and commits as recently as 2026-09-16 including a same-day fix for a reported authorization vulnerability (https://github.com/Helicone/helicone/pull/5816), which is a better signal of an actively-maintained security posture than most vendor 'security page' copy.

Helicone was acquired by Mintlify (the AI-native docs platform) on 3 March 2026; the deal was announced on Helicone's own blog, which states the product will stay live 'in maintenance mode' — meaning security updates, new model support, and bug fixes continue, but the company is no longer operating as an independent, actively-growing startup (https://helicone.ai/blog/joining-mintlify). Y Combinator's company page also lists Helicone's status as 'Acquired' (https://www.ycombinator.com/companies/helicone). In practice this has not shown up as a slowdown on GitHub — the main repo has 25+ merged PRs since the acquisition date and a security fix shipped 2026-09-16 — so 'maintenance mode' has held up as a description rather than a euphemism for abandonment, at least so far.

The tagline's 'AI gateway' component lives in a separate GitHub repository, Helicone/ai-gateway, not Helicone/helicone. Two things differ from the main platform: the licence, and the pace of work. The gateway repo was relicensed from Apache-2.0 to GPL-3.0 on 2025-11-21 (https://github.com/Helicone/ai-gateway/commit/9649b27bdc9fb0907d359e899894102a15f3a085) — GPL-3.0 is copyleft, a materially different obligation for anyone embedding or modifying the self-hosted gateway than the permissive Apache-2.0 licence the observability platform still carries. That relicensing commit is also the last commit in the repository: no code has landed since (repo pushed_at is 2025-11-21, per the GitHub API), an open feature PR from June 2026 sits unmerged (https://github.com/Helicone/ai-gateway/pull/305), and the npm package's latest published version, 0.2.0-beta.30, is from July 2025 — it never even picked up the GPL relicense in a new release. The repo isn't archived and the cloud-hosted gateway endpoint (ai-gateway.helicone.ai) is live and still gets model additions through the main platform repo, so this reads as a component the team stopped actively developing in the open rather than a shut-down product — but a buyer evaluating the self-hosted, open-source gateway specifically should know it's frozen at a beta build from over a year ago under a licence that changed after most of the code was written.

Hobby (free): 10,000 requests/month, 1GB storage, 1 seat. Pro: $79/month, unlimited seats, alerts, reports, and HQL, plus usage-based charges above the included quota. Team: $799/month, adds SOC-2/HIPAA compliance docs, 5 organizations, a dedicated Slack channel. Enterprise is quote-based and adds SAML SSO, custom MSAs, and on-prem deployment. All four tiers and the usage-based calculator are published on the pricing page with a live cost estimator (https://helicone.ai/pricing, checked 2026-09-24). Startups under 2 years old with under $5M raised get 50% off the first year; open-source projects get a $100 credit; students get the paid tier free — all stated on the same page.