4 Helicone Alternatives to Migrate To Before It Freezes (2026)

If you’re on Helicone, you need a migration plan. Not today, not in a panic, but on the calendar.

Here’s why. Mintlify acquired Helicone on 3 March 2026 and put it in maintenance mode. Only security patches, bug fixes and new-model support ship now. No new features, no new integrations, no roadmap. And Mintlify has said it’s actively helping existing customers migrate off - which is a rare thing for an acquirer to admit out loud. When the company that owns your observability tool is pointing you toward the door, the tool is a dead end no matter how clean it is.

So this isn’t a “Helicone was bad” list. It wasn’t. It was a tidy, open-source, proxy-based logger that processed 14.2 trillion tokens across roughly 16,000 organizations before the acquisition. This is a list for the one reason everyone’s here now - the product is frozen - and I’ve matched each alternative to why you picked Helicone in the first place. That’s the part that decides your migration. Move by the job Helicone did for you, not by whichever tool ranks highest on some generic list.

The short version

Tool	Best for	Starting price	Self-host	License
Helicone	Nobody new - existing users planning an exit	$79/mo	Free, frozen codebase	Apache 2.0
Langfuse	Open-source observability you can self-host free	Free / $29/mo	Free, full product	MIT
Opik	Cheapest managed cloud, permissive OSS	Free / $19/mo	Free, no gates	Apache-2.0
Portkey	The gateway and proxy angle Helicone had	Free / $49/mo	Gateway free, observability paid	Apache 2.0 gateway
LangSmith	Turnkey, well-funded, LangChain apps	$39/seat/mo	Enterprise only	Closed

If you left because you want open source that’s still alive: Langfuse

Helicone’s best trait was that it was genuinely open source - Apache 2.0, self-hostable via Docker or Helm with no license fee. The problem now isn’t the license, it’s that the codebase is frozen. Self-hosting an open-source tool is a bet that upstream keeps improving and you benefit for free. With Helicone, upstream is done.

Langfuse is the same bet on a project that’s still moving. It’s the open-source observability default, MIT-licensed, and self-hosting is genuinely feature-complete - only three features are enterprise-gated, so tracing, evals, prompt management, human annotation and RBAC are all free when you run it yourself. Framework-agnostic, actively developed, and dramatically cheaper than the closed alternatives at scale - Langfuse’s own figures put it around $101/mo at 1M events versus roughly $2,514/mo for LangSmith at comparable volume.

The honest catch is operational. Langfuse v3 moved from “just Postgres” to Postgres plus ClickHouse plus Redis plus S3-compatible storage - four services to stand up, and the migration to that architecture is where self-hosters get stuck, with a string of GitHub threads reporting memory errors and timeouts. If your Helicone self-host was a single container, this is a step up in ops weight. One more fact to file: Langfuse was acquired by ClickHouse in January 2026, so you’re trading one acquired tool for another - the difference is that ClickHouse committed to keeping the MIT license and self-host, and the roadmap is still shipping. Details are on the Langfuse page.

If you left because you want a cheap managed dashboard, not ops: Opik

Plenty of Helicone users were on the free Hobby tier or the $79 Pro plan precisely because they didn’t want to run infrastructure. If that’s you, don’t replace a frozen managed tool with a four-service self-host. Replace it with a cheaper managed tool.

Opik is Comet’s open-source observability and eval platform, and its Pro cloud tier is $19/mo for 100k spans - the cheapest paid cloud of the major eval platforms, with overage at $5 per additional 100k spans and a free tier at 25k spans a month. The nice part is the optionality: if you ever do want to self-host, the OSS build is Apache-2.0 with the full feature set and no gates, so you’re not locked into the cloud the way Helicone’s opaque “usage-based” overage left you unable to even forecast the bill. Opik is also the fastest-growing project of its peers at roughly 20.8k GitHub stars, per its own repo.

The honest gotcha is per-seat pricing at scale. The recurring complaint is that seat costs add up as your team grows - the $19 headline is the small-team configuration, not the large one. A few users also report occasional UI slowdown on very large projects. Neither undercuts the core value, but if you’re a big team, model the seat cost before you commit.

If you left because the gateway was the whole point: Portkey

A lot of people used Helicone as a gateway, not an eval tool. You changed your base URL, it routed and logged your calls, done. If that proxy-in-the-middle model is what you actually want to keep, Portkey is the closest continuation - it’s the same drop-in-proxy idea built out into a real gateway that routes to 1,600+ models with fallbacks, load balancing, retries, caching, budgets and 50+ guardrails, and the open-source gateway self-hosts free under Apache 2.0. Point your existing OpenAI SDK base URL at it and you’re routing, no re-instrumentation - exactly the zero-code convenience Helicone sold.

It also improves on Helicone’s biggest structural weakness. Helicone’s proxy sat in the hot path as a single point of failure, and Portkey’s whole reason for existing is routing and fallbacks, so it’s built to fail over instead of just failing.

The catch is that observability is gated. The free open-source gateway gives you routing and a basic dashboard - the real logs, traces, analytics and retention live on the managed Production tier at $49/mo for 100k logs. And that meter caps logs, not requests: past the cap your calls still route, you just stop recording, with overage at $9 per 100k requests. So self-hosting Portkey gets you the pipe, not the dashboard. Portkey also markets sub-1ms added latency, but a competing Kong-run benchmark disputed that - both are interested parties, so test on your own path. The full split is on the Portkey page.

If you left because you want turnkey and vendor-backed: LangSmith

Some teams reach for an alternative and realize the real lesson from the acquisition is that they don’t want to bet on a small, acquirable project again. If you want the opposite of a frozen tool - turnkey, actively developed, and backed by a company that isn’t going anywhere - LangSmith is the most conservative pick. LangChain, Inc. raised a $125M Series B in October 2025 at a $1.25B valuation, per its own funding disclosures, and the tracing is the deepest available for LangChain and LangGraph code. You add a callback and every chain and tool call shows up traced with zero config.

The gotchas are price and lock-in, and they’re real. LangSmith is fully closed source with no free self-host - self-hosting is Enterprise-only, so unlike Helicone you can’t run it yourself to escape the meter. And the meter is steep: base traces are $2.50 per 1,000, which works out to roughly $2,514/mo at 1M base traces on one seat by LangSmith’s published rates, around 25x Langfuse at comparable volume. The deepest value also binds you to the LangChain stack. It’s the turnkey, stable choice - just the most expensive and most locked-in one. The trade-offs are laid out on the LangSmith page.

So which one?

• You valued Helicone’s open-source self-host and want that on a living project - Langfuse, as long as you can run a four-service stack.
• You were on the managed tier and just want a cheap dashboard without ops - Opik at $19/mo, no contest.
• You used Helicone mainly as a gateway - Portkey, knowing the real observability is on the paid tier.
• You never want to migrate off an acquired tool again and will pay for stability - LangSmith if you’re on LangChain and can absorb the bill.

One honest note. There’s no reason to rush a bad migration - Helicone still logs your calls today, security patches still ship, and nothing breaks the moment you read this. But the roadmap is over and the vendor is helping people leave, so the runway only gets shorter. Pick the tool that matches the job Helicone did for you, move deliberately, and don’t replace a frozen product with a heavier one you’ll regret. The figures above are from each vendor’s public pages, verified on 23 July 2026, and this category ships breaking changes monthly - we re-verify every 30 days.