Skip to content
FenixIAm by AIworks 2028 FenixIAm

Platform

One engine underneath every agent. You choose the model; Fenix measures it.

Models from seven providers, real-time voice, measurement of every conversation and data isolated per company. Your agent does its job; the engine makes sure it answers, can be measured and does not overspend.

Engine spec sheet

direct AI providers
7
models ready for your agent
35+
real-time voice engines
4
voices for your characters
≈ 80
Based on production as of 3 October 2026.

01 · Models

The best models, without marrying any of them.

Fenix works with 7 direct providers and has more than 35 models ready for your agent. We do not hand you a list of hundreds: every model we offer has a valid key and works today. If your case needs another one, we connect it through an aggregator with hundreds of models. Each agent can switch model without touching its instructions, its voice or its data.

7 direct providers

  • OpenAI
  • Anthropic
  • Google
  • Mistral
  • DeepSeek
  • Groq
  • xAI

Playground comparator

Example

The same question to three models, with the answers side by side.

ModelLatencyCost
Model AReasoning 3.2 s€0.0140
Model BBalanced 1.4 s€0.0040
Model CFast 0.5 s€0.0008
Illustrative example: latency and cost are not real measurements.
Interchangeable

You change the model, not the agent

More than 35 models available. Moving from one to another means picking it in the settings: the agent keeps everything else.

No broken options

Only what works today

We only offer providers with a valid key. If a model is on the list, it is available to your agent.

Router

The right model for each task

The router picks the model for the task and, if the chosen one does not answer, switches to the backup automatically (failover).

Task Router Chosen model if it does not answer Backup

02 · Voice

Real-time voice, with the voice each character calls for.

Agents listen and speak in real time. Each character has its own voice: the receptionist does not sound like the demanding customer or the doctor in the rehearsal.

4 real-time voice engines

≈ 80 voices to choose from

Voice casting

Example

  • Receptionist Warm, unhurried voice
  • Demanding customer Deep, direct voice
  • Doctor Calm, precise voice
Illustrative example of how a voice is assigned to each character.

Telephony with automatic cut-offs

Each call ends by itself when it reaches its maximum duration or when the line goes silent. A forgotten call does not keep running up costs.

By duration

Cut off at the maximum duration

By silence

Cut off after a long silence

03 · Measurement

Every conversation leaves clear accounts.

What an agent does can be scored, checked and paid for knowing what it costs. These are the instruments.

  1. Reports by criterion

    Each company defines its criteria. The report gives a score per criterion with the sentence that justifies it.

  2. Accuracy with citations

    Each technical claim is checked against your documentation: correct, incorrect or unverifiable, with the citation from the document.

  3. Team roll-ups

    Results aggregated by team, with a PDF report.

  4. Cost of every message

    Every message is measured in tokens and in euros. You know what each agent costs.

  5. Monthly cap per agent

    Each agent has a monthly spending cap. When it is reached, spending stops.

Accuracy with citations

Example

  • “The warranty covers two years.”

    Correct Terms of sale · section 3
  • “Installation is included.”

    Incorrect Current price list · p. 2
  • “It is the best seller in the sector.”

    Unverifiable No source in your documentation
Illustrative example with fictitious documents.

Monthly cap · sales agent

Example

€38.40 of €50.00 this month

Last message €0.0031 · 1,240 tokens

When it reaches €50.00, the agent stops spending.

Illustrative example.

04 · Multi-company One engine, many businesses

A new business is a new page.

Very different businesses run on the same engine, each with its own isolated data, its own brand and its own agents. Adding a new one means configuring it, not building it from scratch.

  • Data isolated per company
  • Your own brand
  • Full white label for public-sector case files
Scene created with AI from our mascot, the paper phoenix.

White label · public case file

Town Council of Example Its name and its coat of arms

  • On every screen
  • On every document
Illustrative example.

05 · Knowledge

Your agents answer with your documents.

Manuals, product sheets and procedures: the agent searches them before answering. And the instructions that work are saved so nobody reinvents the wheel.

Your documents

Search across your documents

Agents search your company’s documents to answer with your information, not with generalities.

Library

Shared instructions

A library of instructions (prompts) that any agent can reuse. Nobody starts from a blank page.

Document search

Example

What is the return period for business customers?

  1. Commercial terms.pdf · p. 7

    Returns from business customers are accepted for 30 days from delivery.

  2. Service manual.docx · section 3.2

    Before accepting a return, confirm the order number and the condition of the product.

Illustrative example with fictitious documents.

06 · Security

Baseline security, no small print.

These are the protections running in production today. We describe them exactly as they are.

  • HttpOnly

    Session in an HttpOnly cookie

    The page’s code cannot read the session, so a third-party script cannot steal it.

  • CSRF

    CSRF protection

    Requests that change data carry a check that stops another website from acting on your behalf.

  • 5 attempts

    Account lockout

    After 5 failed sign-in attempts, the account is locked.

  • Limits

    Usage limits

    Usage limits curb abuse and unexpected spikes.

  • AES-256

    Encrypted keys

    AI provider keys are stored encrypted with AES-256.

  • Consent

    Consent per person

    Each person decides whether to share their results. Without their permission, they are not shown under their name.

Custom-built

Is your process unlike any other? We build it on this engine.

We start from how you work today and design the agent with everything on this page: models, voice, measurement, brand and security.

How we build custom agents

We show you the platform with a case like yours.

In a thirty-minute demo you see the models, the voices and the measurement working, and what it would cost to launch.