Use cases

Your data classified and sorted without leaving your application

Mail, documents, requests: every piece of data entering your software is filed, linked and prioritized the moment it arrives. Your users find everything in its place, with urgent items on top.

Three steps, inside your software

  1. 1

    Your categories

    You define your categories, your folders and your urgency levels, with a few examples.

  2. 2

    Classification

    Every piece of data that arrives is classified immediately.

  3. 3

    Action

    It is filed and linked to the right account or folder, and the right person is alerted if it is urgent.

Human in the loop

Uncertain cases are escalated

When a document or a request could go to two places, the user sees it and chooses. Everything else files itself.

Reflex, Agora's System One decision engine

A System One decision engine answers a precise question about a text or an image in a single, very fast pass of the model: it returns a typed answer (a category, a score, yes or no) along with its probability.

System One · Reflex

Decide

One question, one typed answer and its probability, in a single pass. For short, repeated decisions: filing a message, choosing a folder, setting an urgency level.

Large language model

Reason

Writing, chaining steps, untangling an unusual case. It steps in when the decision calls for thought.

In your agents, Reflex handles the day's short decisions, and the large language model focuses on what requires reasoning.

1 · Accuracy

A reliable decision, or a question

Every decision comes with its confidence level. When it is too low, the case goes back to the user instead of being filed at random.

2 · Volume

The load of all your clients

Throughput guaranteed by contract, at full load. Requests above it are flagged so they can be retried automatically: overages are never billed.

3 · Cost

A bill that no longer tracks volume

Each decision made by Reflex replaces a text generation with a single pass of the model. And your monthly cost is fixed, whatever the number of decisions.

Is your application already agentic?

Speed up your workflows without touching your harness

Your agents chain short decisions: route, choose, validate. Handing them to an LLM slows down every step and drives up the bill. Plug Reflex into your harness: your agents decide fast, and keep the LLM for what requires reasoning.

Capacity-based licensing

You pay for peak throughput, with no per-token billing

Your subscription sets a capacity tier, expressed in decisions per second, with throughput and latency guaranteed by contract. Your monthly cost stays fixed as long as your peaks remain within your tier, and you move up a tier as your usage grows.

One call, one actionable answer

Reflex is called via API, or via MCP like any other tool in your agent. The question applies to a text or an image, and a single call can ask several questions about the same content. The answer is typed, with its probability: your code uses it directly, with no free text to interpret.

Compatible with the System One model API contract: if you already use one, you switch to Reflex without changing your code, and you gain the ability to send an image in the question.

On your harness or on ours

Your harness

Your classification agent runs on your own harness and calls Reflex via API, or via MCP as a tool.

Agora harness

Agents built on the Agora platform use Reflex natively.

Show us what your users receive

A real example of mail, a document or a request: we discuss it in a first conversation.

What do you want to classify?
How many items per day, across all clients?
How is it done today?
Do you have your own agentic harness?

Your details are only used to get back to you about this request. Terms of service

Request a demo