Is ClassifierHub cheaper than the OpenAI Decisions API?

For long inputs, yes: a credit costs the same whatever the length, and on the Business plan it beats OpenAI's per-token price above roughly 2,500 to 3,000 input tokens. For short text, OpenAI's raw per-call price is lower, but you build and run the surrounding tooling yourself.

Can I use ClassifierHub from Claude, Cursor or other AI tools?

Yes. ClassifierHub ships a remote MCP server and a local npx package, so any MCP client can call classify, score, decide and batch tools with your API key.

Does ClassifierHub use OpenAI models?

No. ClassifierHub runs on TypeSafe's Jev decision model through OpenRouter today, and its engine is provider-agnostic. It is not affiliated with OpenAI, TypeSafe or OpenRouter.

Can ClassifierHub classify images like OpenAI's Decisions API?

Not yet. ClassifierHub accepts text and JSON. If image decisions are essential today, OpenAI's endpoint supports base64 images.

How do I try ClassifierHub?

Create a free account: the Free plan includes 100 credits a month in the playground. API, MCP and batch access start on the Starter plan at €9 a month, currently 50% off at launch.

OpenAI Decisions API alternative

ClassifierHub vs. the OpenAI Decisions API

Short answer

OpenAI's Decisions API is a beta endpoint billed per input token. ClassifierHub is a decision layer with a flat price per decision, up to five answers per credit, plus an MCP server for Claude, Cursor and other agents, Excel and CSV runs, templates and batch jobs. It's cheaper for long inputs and needs no plumbing.

OpenAI's Decisions API validates something we've built ClassifierHub around: most decisions inside software don't need a chat model that writes paragraphs. They need a fast, typed answer with a probability you can act on.

The two products overlap in that idea but differ in what you get. OpenAI ships a raw endpoint on one model. ClassifierHub ships the decision plus everything around it, usable from code, AI agents and spreadsheets on day one. Here is a fair comparison, including where OpenAI comes out ahead.

Get Early Access to ClassifierHub: 2× credits in your first paid month.

Side by side

	ClassifierHub	OpenAI Decisions API
Pricing model	Flat credits: 1 credit per decision, up to 5 answers	$0.10 per 1M input tokens, output free
Cost grows with input size	No, same credit for a sentence or a long document	Yes, plus regional and long-context premiums
Question types	Choice, yes/no, score; templates and saved decisions	Predicate, choice, score
Inputs	Text and JSON	Text and base64 images
MCP server for AI agents	Yes, remote and local (Claude, Cursor, any MCP client)	Not part of the endpoint
Excel and CSV runs	Built in	Build it yourself
Ready-made templates	Yes: support and email triage, lead qualification, spam, agent guardrails	No
Batch jobs, usage and credit budgets	Built in	Build it yourself
Models	Provider-agnostic; saved decisions keep working if the model changes	gpt-6-luna only
Status	Available, with 50% launch pricing	Public beta

Which one is cheaper? The honest math

For short text, OpenAI's raw price per call is lower. A 300-token support message costs about $0.00003 on the Decisions API. A ClassifierHub decision costs between €0.0009 (Starter, €9 for 10,000 credits) and €0.000245 (Business, €49 for 200,000 credits), VAT included.

The picture flips as inputs grow, because a ClassifierHub credit doesn't depend on length. On the Business plan the break-even is around 2,500 to 3,000 input tokens per decision, depending on the exchange rate. Email threads, contracts, tickets with history, product records and agent transcripts often sit above that, and OpenAI's regional processing premiums (for EU data residency, for example) and long-context multipliers move it further in ClassifierHub's favor.

Price per call is also not the whole bill. With a raw endpoint you write and maintain the prompt templates, retries, batch runner, usage tracking, API keys per team, spreadsheet import and an MCP server for your agents. With ClassifierHub those are included in the plan, and the monthly price is fixed, which makes budgets predictable.

Plans start free. Paid plans are 50% off during launch.

See pricing

Works in every AI tool you use, not just your code

ClassifierHub exposes the same decisions through a REST API, a remote MCP server, a local stdio MCP server and an Excel and CSV runner. Add it once to Claude, Cursor or any other MCP client and your agents get classify, score, decide and batch tools, with the same credits and limits as the API. Anything that can make an HTTPS request, such as Zapier webhooks, n8n or Apps Script, can call it too.

OpenAI's Decisions API is reached through OpenAI's API and SDKs. That's ideal inside an OpenAI-based backend, but an agent in another client or a teammate working in a spreadsheet needs something built on top.

mcp.json

json

{
  "mcpServers": {
    "classifierhub": {
      "url": "https://classifierhub.com/api/mcp",
      "headers": { "Authorization": "Bearer ch_live_..." }
    }
  }
}

The ClassifierHub API opens to Early Access members first.

Give your agents decision tools in under a minute.

Set up MCP

Reliability and lock-in

ClassifierHub separates your decisions from the model that answers them. Templates and saved decisions define the questions and options; the engine retries transient provider errors with backoff and is built to switch providers without changing your integration. If a model is deprecated or repriced, your API calls, MCP tools and spreadsheets stay the same.

OpenAI's endpoint is in beta and currently supports one model. That's a fine starting point, but plan for request or pricing changes at general availability.

Migrating takes minutes

The request shapes map one to one. A choice question becomes a classify call: OpenAI's choices with value and description become ClassifierHub options, and instructions stay the same. A predicate becomes a yes/no decision and a score becomes a score call with ordered levels.

classify.sh

bash

curl https://classifierhub.com/v1/classify \
  -H "Authorization: Bearer $CLASSIFIERHUB_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "input": "I was charged twice for my order.",
    "instructions": "Which department should handle this complaint?",
    "options": {
      "billing": "Payments, invoices, and refunds.",
      "technical": "Problems using the product.",
      "other": "Requests outside these categories."
    }
  }'

# -> { "label": "billing", "confidence": 0.93,
#      "probabilities": { "billing": 0.93, "technical": 0.04, "other": 0.03 },
#      "usage": { "credits": 1, "latency_ms": 420 } }

The ClassifierHub API opens to Early Access members first.

When to pick OpenAI instead

Choose OpenAI's Decisions API if you need image inputs today, your decisions are on short text at very high volume and you're optimizing purely for raw price per call, or your team is committed to a single OpenAI stack and is happy to build the tooling around the endpoint. Choose ClassifierHub if you want decisions available to agents, spreadsheets and code at once, long inputs at a flat price, and no infrastructure to maintain.

Frequently asked questions

Related guides

OpenAI Decisions API

OpenAI Decisions API: what it is, what it costs and its limits

Comparison

Jev vs. GPT-style LLMs for classification

LLM cost reduction

How to reduce LLM costs without hurting quality

Last updated 2026-10-08. ClassifierHub is an independent product built on top of the Jev decision model, accessed through OpenRouter. It is not affiliated with or endorsed by TypeSafe or OpenRouter.