Portkey vs Helicone in 2026 - Why This Comparison Already Has a Winner

If you are comparing Portkey and Helicone, you are usually asking one question - which proxy should sit in front of my LLM calls to route them, watch them, and control the spend. Both products answer that. Both work by putting themselves in your request path. On the surface it looks like a close fight.

It is not close anymore. Helicone went into maintenance mode after Mintlify acquired it on 3 March 2026 - security patches, bug fixes and new-model support only, no roadmap, and Mintlify is actively helping customers migrate off. For a new buyer that single fact ends the comparison. So this post does two things: it explains why Portkey wins by default, and it is honest about Portkey’s own catch so you know what you are actually buying. I have also pulled in Langfuse, because “gateway plus proxy” is often not what you actually needed.

The short version

Tool	What it leads with	Self-host	Real observability	Starting price
Portkey	LLM gateway - routing, fallbacks, budgets	Gateway only, Apache 2.0	Managed tier only	Free / $49/mo
Helicone	Proxy observability - frozen, maintenance mode	Full, Apache 2.0, but frozen	Yes, but no roadmap	Free / $79/mo
Langfuse	Open-source observability and evals	Full, MIT, free	Yes, full-featured	Free / $29/mo

Portkey: the gateway that happens to include traces

Portkey is an LLM gateway. It sits between your app and your providers as a drop-in proxy and routes calls to 1,600+ models through one API, adding fallbacks, load balancing, retries, caching, budgets, rate limits and 50+ guardrails. The pitch is spend and routing control across many providers, not eval depth. If your problem is “we call five providers, costs are exploding, and we need fallbacks and budgets in one place,” Portkey is aimed squarely at you. One base URL change routes your existing OpenAI SDK calls through it, no re-instrumentation.

It is well backed - founded in January 2023 by Rohit Agarwal and Ayush Garg, with a $3M seed and a $15M Series A led by Elevation Capital. The open-source gateway sits around 12,500 GitHub stars.

The gotcha is where the value migrates. The Apache-2.0 gateway self-hosts free, but that gets you routing and a basic dashboard - not real observability. The logs, traces, analytics and retention that most teams adopt an observability tool for live on the managed Production tier at $49/mo for 100,000 logs. And the meter caps logs, not requests, at $9 per additional 100k requests - so past the cap your traffic keeps routing while your observability quietly goes dark. One more caveat: Portkey markets sub-1ms added latency, but a benchmark run by Kong, a competing gateway vendor, reported Portkey at roughly 65% higher latency than Kong’s gateway. Both are interested parties, so test on your own path.

Helicone: clean product, dead-end status

Helicone was the lightest-weight observability tool of the bunch - a logging proxy first, evals a distant second. You change your API base URL to route through it and it logs requests, tracks cost, and gives you a dashboard and a router across providers. It works with any language and provider through the proxy, with an async SDK logging mode for when you do not want the proxy in your critical path. It came out of Y Combinator, and by the time Mintlify bought it, it had processed 14.2 trillion tokens across roughly 16,000 organizations.

None of that matters for a new buyer, because the product is frozen. Mintlify put Helicone into maintenance mode on 3 March 2026 - security and bug fixes and new-model support only, no new features, no roadmap, and Mintlify is pointing customers toward the exit. The Apache-2.0 self-host still works and is more open than Portkey’s split model, but self-hosting a codebase whose own owner is winding it down is a weak bet.

There is a structural gotcha underneath the acquisition too. The proxy model is a single point of failure. In proxy mode Helicone sits in the hot path of every LLM call - if Helicone is down, your calls fail even when OpenAI and Anthropic are healthy, and every proxied call adds latency (Helicone cites around 10ms). The async SDK mode dodges both problems but costs you the zero-code convenience that was the main reason to pick Helicone.

The third option: Langfuse, if you wanted observability all along

Here is the honest reframe. A lot of people land on “Portkey vs Helicone” because they want tracing and cost visibility, and both happen to offer it. But if observability is the actual goal - not routing - Langfuse is the stronger pick than either. It is MIT-licensed and self-hosts free with the real product intact - only three features are enterprise-gated. At 1M events a month it runs about $101/mo managed, and self-hosting removes the per-trace cost entirely. It is framework-agnostic and runs as an OpenTelemetry backend.

Langfuse has its own catch, and it is operational. The v3 self-host needs Postgres plus ClickHouse, Redis and S3-compatible storage - four services - and the migration to that architecture is where self-hosters get stuck. It is also a ClickHouse subsidiary after the January 2026 acquisition. But it is actively developed, genuinely open, and does not put itself in your request hot path the way a proxy does.

Portkey vs Helicone: which should you pick?

• You need routing, fallbacks, budgets and guardrails across many providers - Portkey. It is the strongest gateway in this set, and the free self-hostable core is genuinely useful. Just plan for the managed tier the moment you need real logs.
• You are building a new project and want observability, not a gateway - skip both proxies and start on Langfuse. Free self-host, full features, no request-path risk.
• You are already on Helicone today - use the runway to plan your migration. The product is clean but frozen, and its owner is helping people leave.
• You want the absolute lightest cost-tracking proxy and accept a frozen tool - Helicone still technically does that, but I would not start there in mid-2026.

The honest summary: this is a rare comparison where the answer is settled by a corporate event, not a feature. Portkey is a real, actively developed gateway with a clear paid-observability catch. Helicone is a tidy product its own owner has stopped building. If you want a gateway, Portkey. If you want observability, Langfuse. Helicone is the answer to neither for anyone starting fresh.

Every price and date here was read from each vendor’s own materials and verified on 26 July 2026. This category ships breaking changes monthly, so we re-verify every 30 days.