Built on Jev, TypeSafe's decision model, via OpenRouter
Give your AI agents a fast decision layer.
Classify, score, route and validate information with a fast, low-cost AI decision layer before sending expensive reasoning tasks to larger models.
Input
“Hi, our card was charged twice for the September invoice. Can you refund the duplicate? Our finance team closes the books today.”
Decision
The problem
Most AI calls in your stack are simple decisions.
Large AI models are powerful, but they're slow and expensive for questions that only need a category, a yes or a no, or a score. Teams end up either paying a reasoning model to sort their inbox, or doing it by hand.
These decisions shouldn't need expensive reasoning. They need to be fast, cheap, structured and reliable.
- Should this email be escalated?choice
- Is this lead qualified?yes / no
- Is this message spam?score
- Which team should receive this request?choice
- Does this agent action need human approval?yes / no
- Is this document an invoice?score
How it fits
A decision layer between your data and your actions.
Put ClassifierHub in front of your workflows. It decides in a single call and hands your code a clean, structured answer.
01
Input
An email, lead, ticket, form, document row or agent action.
02
Fast decision layer
Your decision runs on a small, fast decision model.
03
Structured result
Typed answers with probabilities: never free text to parse.
04
Action
Route it, automate it, or escalate to a larger model only when needed.
{
"id": "exec_7f3a9c...",
"outputs": {
"category": { "type": "choice", "value": "billing", "confidence": 0.94 },
"urgent": { "type": "boolean", "value": true, "probability": 0.81 },
"sentiment": { "type": "score", "value": 1, "label": "Neutral" }
},
"usage": { "credits": 1 }
}What you save
Save hours of manual work, and stop overpaying for simple AI calls.
Time savings
Automate the repetitive decisions your team makes all day, so people only handle what actually needs a person.
- Email classification
- Lead qualification
- Support routing
- Contact form processing
- Document classification
- Repetitive checks
AI usage savings
Handle classification, routing and yes/no checks with a cheaper decision layer, and reserve expensive reasoning models for the tasks that actually need them.
Without a decision layer
- Every request
- Large reasoning model
With ClassifierHub
- Every request
- Decision layer
- Only complex cases → large model
Actual savings depend on your workload. We'll publish benchmarks, not promises.
Use cases
What you can automate
Start from a ready-made classifier or describe your own decision. Every result comes back structured.
Email classification
Stop reading every email to decide where it goes. Route by intent and flag what needs a reply today.
- Gmail / inbox
- Decision layer
- Routed + labeled
Example input
{
"from": "ops@acme-logistics.com",
"subject": "Invoice #4821 charged twice",
"body": "Our card was charged twice for September..."
}Structured result
{
"category": {
"value": "billing",
"confidence": 0.94
},
"urgent": {
"value": false,
"probability": 0.21
}
}Illustrative example of the response format.
Stop routing work by hand.
Early Access members get 2× credits in their first paid month.
How it works
From idea to automated decision in five steps.
- 1
Describe what you want to decide
Pick a template or define outputs: categories, yes/no questions or scores.
- 2
Test the decision
Paste real examples in the playground and check the probabilities.
- 3
Connect your data
Send it from your app, an AI agent, a spreadsheet or an integration.
- 4
Use the API, MCP or an integration
One call, one structured answer. Same decision everywhere.
- 5
Automate the result
Route, label, escalate or hand off to a larger model when it matters.
For developers
Coming soonOne API call. A typed decision back.
A versioned REST API with consistent JSON, clear error codes and credits reported on every response. MCP for AI clients. SDKs later.
Architecture
- Your app
- Decision API
- Decision engine
- Jev
The decision engine is provider-agnostic. Jev is the first model; more can be added without changing your integration.
curl https://classifierhub.com/v1/classify \
-H "Authorization: Bearer $CLASSIFIERHUB_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"input": "My invoice was charged twice",
"instructions": "Which team should handle this?",
"options": ["sales", "support", "billing"]
}'const res = await fetch("https://classifierhub.com/v1/decide", {
method: "POST",
headers: {
Authorization: `Bearer ${process.env.CLASSIFIERHUB_API_KEY}`,
"Content-Type": "application/json",
},
body: JSON.stringify({ template: "support-triage", input: ticket.body }),
});
const { outputs } = await res.json();
if (outputs.escalate.value) await pageOnCall(ticket);{
"mcpServers": {
"classifierhub": {
"command": "npx",
"args": ["-y", "@classifierhub/mcp"],
"env": { "CLASSIFIERHUB_API_KEY": "ch_live_..." }
}
}
}MCP
Coming soonDecision tools for Claude, Cursor and your agents.
Connect the decision layer to Claude, Cursor and other AI tools through MCP. Your agent asks a quick, cheap question instead of reasoning through it.
Example
- Claude
- Gmail MCP
- ClassifierHub MCP
- Decision
- Action
Integrations
Use it where your data already lives.
We label every integration honestly: what works today, what's next, and what's planned.
Launch pricing
Simple credit-based plans. 50% off at launch.
Early Access members get 2× credits during their first paid month.
Free
100 credits / month
Starter
50% launch discount10,000 credits / month
Early Access first month
20,000 credits (2×)
Pro
50% launch discount50,000 credits / month
Early Access first month
100,000 credits (2×)
Business
50% launch discount200,000 credits / month
Early Access first month
400,000 credits (2×)
FAQ
Questions, answered.
Reserve your Early Access.
Join the waitlist today and get 2× credits during your first paid month when we open your spot.