Is Helicone shutting down?

Helicone says its services remain live for the foreseeable future in maintenance mode: security updates, new models, and bug fixes.

Can ReqKey cap a user’s AI spend?

Yes. Give the user a credit balance, charge each action what it costs you, and requests stop at zero until they refill or top up.

Does ReqKey count tokens?

You set the price. Charge a flat cost per route, or check the key before a stream starts and settle the cost from the final token count when it ends — the two-phase pattern in the AI metering guide.

ReqKey vs Helicone · checked September 2026

A Helicone alternativethat meters your users, not your LLM bill.

Helicone went into maintenance mode after Mintlify acquired it in March 2026. ReqKey is actively developed and meters the other direction: your customers’ calls into your AI product.

Start building free See ReqKey pricing

A source for every claim
Where Helicone wins, too
Free for 5M requests a month

The short answer

Helicone proxies your calls to OpenAI and Anthropic to trace prompts and costs. ReqKey sits in your own API and gives each of your users a key, a credit balance, and a rate limit — so one user can’t burn through your model budget.

Helicone is a LLM gateway and observability. ReqKey is API keys, credits, rate limits, and request analytics for the API you sell.

ReqKey, in one line

Free for 5M requests a month. $25 for 10M. No per-seat fees.

See every tier

ReqKey and Helicone, line by line.

A filled mark shows who is stronger on that line; a dash means they’re comparable. Their wins are marked too.

ReqKey

Helicone

How it runsReqKeySDK middleware inside your API — no proxy, no DNS changeHeliconeProxy in front of LLM providers (change the base URL)
API keys for your customersReqKeyYes — prefixed and hashed, one customer can hold many keysHeliconeNo — it meters your LLM calls, not your API
Credits and quotasReqKeyA balance per customer, any cost per call, refills every hour, day, week, or month, optional overageHeliconeRate limits by request count or cost per user or property
Rate limitsReqKeyPer customer, set on the plan, windows from 1 second to 24 hours, 429 with Retry-AfterHeliconeCustom limits on LLM traffic
Request analyticsReqKeyEvery request — endpoint, status, latency, customer — 30 days on Pro, plus an Analytics APIHeliconeLLM cost, tokens, and latency per user
Free tierReqKey5M requests a month, 1,000 keysHeliconeHobby: 10K requests, 7-day retention
First paid tierReqKey$25/mo for 10M requests and 100,000 keys, volume tiers to 1BHeliconePro $79/mo plus usage
Takes paymentReqKeyNo — ReqKey meters and enforces; you take payment with your own providerHeliconeNo

Helicone details from their own pricing pages and docs, checked September 2026. Sources below.

Why teams pick ReqKey over Helicone.

1

Actively developed

Helicone’s services are in maintenance mode — security fixes, new models, and bug fixes only.

2

Meters the right direction

ReqKey meters your customers’ calls to your API, which is how an AI product charges its users.

3

One credit currency

Charge 1 credit for a chat and 40 for an image, whichever model runs behind it.

4

No proxy between you and the model

Your app talks to the provider directly; ReqKey is asked once per user request.

5

500× the free tier

5M free requests a month against Helicone Hobby’s 10K.

When Helicone is the better pick.

We’d rather you pick the right tool than switch twice.

You only need prompt and response tracing plus cost analytics on your own OpenAI or Anthropic usage.
You accept a frozen roadmap for the features you already use.

Moving from Helicone in three steps.

1
Keep tracing if you need it
Prompt tracing and user metering are different jobs — you can run an LLM observability tool alongside ReqKey.
2
Give each user a key and a balance
Create a consumer per user with a credit allowance and a rate limit from their plan.
3
Charge per action
Add the middleware and price each route — chat, image, embedding — in credits.

Framework guides How ReqKey works

ReqKey vs Helicone: common questions.

Something else? Ask the team or read the docs.

Helicone says its services remain live for the foreseeable future in maintenance mode: security updates, new models, and bug fixes.
Yes. Give the user a credit balance, charge each action what it costs you, and requests stop at zero until they refill or top up.
You set the price. Charge a flat cost per route, or check the key before a stream starts and settle the cost from the final token count when it ends — the two-phase pattern in the AI metering guide.