SYLEX BY AGORA

Quota-free AI
for your teams

Sylex puts local generative and agentic AI directly inside your infrastructure. A fixed budget, no usage-based billing, and code that stays with you by design.

A 30-minute meeting with a live demo, no commitment.

Download the product sheet (PDF, in French)

Your teams want AI. Three things are holding you back.

The appetite is there, and so are the tools. What gets in the way is the terms of use.

Code under lock and key

Your code, and sometimes your customers' code under confidentiality clauses, cannot go to a third party. So many software publishers have ruled out the assistants on the market: their developers work without them while competitors equip themselves.

Usage capped by your contract

Flat-rate or per-token, what your teams can do with AI depends on your contract rather than their needs. Always-on automations are the first to go: one software publisher shelved its commit-review agent because it alone used up 15% of the team's daily quota.

The fear of lock-in

“What if we lose access to this model tomorrow?” Terms that change without notice, tools whose future is uncertain: it is hard to roll out what works on a foundation you do not control.

An inference endpoint inside your infrastructure

Sylex runs on a station installed on your premises (Sylex Station), which gives your teams an OpenAI-compatible API on your own network.

Your developers IDEs, harnesses, CI/CD Sylex Station Local inference OpenAI-compatible API On your premises, on your LAN

Agora tracks new releases and keeps the models continuously up to date, so you do not need to hire MLOps skills. We select the relevant models and stay agnostic about the best frontier model available. Your teams keep their habits: the foundation evolves, not the way they work.

Sylex Frontier, planned for Q2 2027, will add processing capacity operated by Agora on isolated infrastructure, for tasks that go beyond your Station. It changes nothing above: the Station runs on its own, on your premises, and Frontier will remain an option.

A real Station, not a concept

Sylex console: status of a station in production (uptime, model served, disk, license)

Uptime, model served, license, free disk space: the actual status of a station in production, visible from the console.

Sylex vs. the alternatives

Each alternative solves part of the problem. Sylex meets all four criteria that matter to your teams; only the hardware has to be bought.

Sylex Mistral Code DIY
(Ollama, ODS...)
Sovereign APIs
(OVHcloud, Scaleway)
Generic
local box
Fixed price, no usage-based billing
Code stays with you by design
Models kept at the open-weight state of the art
No infra/AI team required
No hardware to buy
Request a demo

30 minutes to see two agents running at the same time on a single Station, with no meter ticking.

What your teams can do with it

A few examples among many, from development to production.

Developers

Writing and reviewing code

Assisted coding, bug detection, getting to grips with legacy code.

Developers

Pull request review

Summarizing and preparing PRs before human review.

Quality

Tests

Faster test coverage on existing or new code, run continuously.

Quality

Code audit

Security and performance, reviewed overnight across the whole repository.

Support

Ticket analysis

Summaries, recurring patterns, incident reports, personal data included.

Product

Customer feedback

Sorting verbatims, reviews and NPS feedback to feed the roadmap.

Product

Documentation

Writing and updating technical and product documentation.

Production

Log monitoring

An agent reads logs continuously and alerts the team on Slack.

Your Station works nights too

With per-token pricing, every task you run adds to the bill. With a fixed price, off-peak hours are already paid for, so you might as well put them to work.

During the day

Your teams write code, review pull requests and have support tickets analyzed.

End of day

Tests run on the day's code, and pull requests are prepared for the next morning's review.

Overnight

An agent reviews the whole repository for security and performance, and leaves its report for the morning.

Around the clock

An agent monitors production logs and alerts the team on Slack.

It all fits in the same subscription: nights cost you nothing more than days.

Works with the harnesses your teams already use

Use Harness
Writing and reviewing code day to day OpenCode
An agent you can reach in Slack, running on your Station Hermes Agent
Your own scripts, your CI, your automations the Sylex API, OpenAI-compatible

These harnesses are qualified by Agora on a Station, with configuration recipes provided in the admin console. Other tools on the market rely on the same interface and should work: tell us which one you use and we will tell you where we stand.

Claude Code speaks a different format and is not supported today. Codex accepts a custom provider; we have not qualified it yet. Either way, your developers keep using them for what they use them for today, and Sylex takes on the background work.

What it changes for you

A budget under control

A fixed, predictable price, whatever your teams' usage volume. No foundation model, no per-token billing.

Sovereignty

Your source code stays with you by design, with no dependency on external services.

Simplicity

Once the box arrives, a single command line and your license key are all it takes. Allow half a day to be up and running, model chosen and harness connected. No AI skills required.

Scalability

What a Station can handle depends on your usage, not on a catalog figure: we size it with you at the start. If your needs grow, you add a Station.

Always at the open-weight state of the art

New open-source models come out every week. Agora keeps them up to date for you, without your teams changing their habits.

Reversibility

The machine stays yours, and your tools speak the market's standard API format. If you change your mind, your scripts and harnesses plug in elsewhere without being rewritten.

Sovereign AI: what it is, and why Sylex is its highest degree

Sylex model catalog: reasoning, code, speed and vision comparison by model

Today's best model, never frozen

Every candidate model (Qwen, poolside, inclusionAI, and those to come) is evaluated by Agora on reasoning, code, agentic use, speed and vision before it enters the catalog. Your teams deploy the one that suits their use in one click; Agora handles the ongoing monitoring and evaluation.

No lock-in to a single model provider: that is how you stay at the open-weight state of the art without depending on one vendor's release schedule.

A console to steer, not to tinker

The Agora admin console lets you choose LLM models, create API keys per application, and monitor your teams' usage in one place. No more patchwork of tools picked individually by each developer.

Sylex admin console: API key management per application

For your CISO, your DPO, your contract clauses

Your code stays with you by design: the Station runs with no dependency on external services. When Sylex Frontier becomes available (planned for Q2 2027), it will be operated by Agora on isolated infrastructure. No code or data is used to train models. Access is controlled through dedicated API keys per application, with an admin console that monitors all usage.

If your customers impose strict confidentiality clauses on code, or if you serve regulated sectors (healthcare, HR, finance, legal, public sector), this is precisely what Sylex was designed for.

A fixed price, whatever the usage

A monthly subscription, independent of your teams' usage volume and number of users. These prices cover the Sylex Software; the Hardware is purchased separately.

Your teams' usage → Cost Per token Flat plan quota reached Sylex +1 Station
Principle, not to scale: how cost evolves with usage depending on the billing model.
Plan 1-year commitment RECOMMENDED No commitment
Sylex Software (per Station) €480/month €600/month

The only limit is the machine. And it belongs to you.

No token meter, no requests-per-minute cap, no tier to unlock with a provider. Your teams use Sylex as much as their Station's capacity allows. If your needs grow, you add another one.

The Hardware is purchased separately, around €5,000 excl. VAT. So expect around €10,800 excl. VAT in the first year, hardware included, then €5,760 excl. VAT per year.

If you stop the subscription, the machine stays yours. You lose access to the inference endpoint and the admin console.

1-year commitment, €480/month

Our standard plan: 20% cheaper than no commitment, in exchange for a 12-month commitment.

No commitment, €600/month

The same offer, paid month by month and cancelable at any time, if you would rather validate usage before committing.

These prices include access to the monitoring console, support, maintenance and updates of the Sylex Software, including updates to the embedded language models. The Station's reference Hardware: Asus Ascent GX10, or any machine with the same Nvidia GB10 architecture (Grace Blackwell, 128 GB, 1 TB) from other manufacturers (Acer Veriton GN100, Dell Pro Max with GB10, HP ZGX Nano AI Station, Lenovo ThinkStation PGX...), purchased separately.

Happy with the price?

Request a demo to see Sylex running on your use cases.

Download the product sheet (PDF, in French)

How to get started

You plug it in, you use it, we maintain it.

1

The machine

You order the box from the manufacturer of your choice. While it ships, we prepare your configuration.

2

Setup

When it arrives, one command line and your license key are all it takes. You pick your model from the console and plug in your harness. Allow half a day.

3

The first agent, then the next ones

A recipe that points at your repository, runs overnight and hands you a result in the morning. You write the next ones yourselves.

Sylex is right for you if...

Team size From a small team to several dozen people.
Sensitive code Critical intellectual property, customers under confidentiality clauses, regulated sectors.
AI already in use Your developers already use cloud tools (Copilot, ChatGPT, Claude Code...) with no central governance.
Budget control Flat-rate or per-token, your teams watch the quota instead of working, and you want a budget that does not cap usage.

Frequently asked questions

How is Sylex different from GitHub Copilot, ChatGPT or other cloud offerings?

Those tools run in their vendor's cloud and are paid by subscription or by usage, with quotas. Sylex runs on a station installed on your premises: your code never goes to a third party, and the price is fixed, with no meter running. Sylex does not aim to replace them everywhere. Your developers can keep their tools for what they use them for today, and hand Sylex the sensitive code and the background work: tests, reviews, audits, overnight jobs.

Can we try the offer before committing?

Yes: it all starts with a demo by one of our experts, on your use cases. Depending on your profile, two trial paths are then available: temporary access to a Sylex Station hosted by Agora, or direct installation of your own Station if your confidentiality clauses require it. Either way, full confidentiality only begins the day the Station is physically on your premises.

What is the return on investment for our teams?

The key argument: a fixed price, with no meter running. Per token, the bill climbs with usage; on a flat rate, the quota caps what your teams can do. With Sylex, heavy usage and always-on agents cost nothing extra: the more your teams use it, the better it pays off.

How are source code confidentiality and security guaranteed?

Your code stays with you by design: the Station runs with no dependency on external services. Frontier, planned for Q2 2027, will be operated by Agora on isolated infrastructure. No code or data is used to train models. Access is controlled through dedicated API keys per application and an admin console that monitors usage.

What is the difference between Sylex Station and Sylex Frontier?

The Station is the foundation: the Sylex Software installed on your Hardware, on your premises. It runs on its own, and we size it with you based on your usage; to add capacity, you add a Station. Sylex Frontier, planned for Q2 2027, will add processing capacity operated by Agora on isolated infrastructure, for tasks that go beyond your Station. It will be an option, not a prerequisite.

Who handles Hardware maintenance and software updates?

The Hardware is purchased separately and remains your property: its physical maintenance falls under the manufacturer's standard warranty. On the software side, Agora handles maintenance and updates of the Sylex Software, including replacing language models, all included in the subscription price.

Is Sylex compatible with our existing tools (IDE, Git, CI/CD, Slack)?

Sylex exposes an OpenAI-compatible API on your network, the format used by AI providers across the market. Two harnesses are qualified by Agora on a Station: OpenCode for code, and Hermes Agent, reachable in Slack. Your scripts and CI call the API directly. Other tools rely on the same interface and should work: tell us which one you use and we will tell you where we stand.

What technical support is included?

Software maintenance and updates are included in the subscription, with access to the monitoring console to manage your teams' usage. Support with the initial deployment is also available.

Ready to give your teams quota-free AI?

Book a 30-minute demo to see Sylex running on your use cases.

Request a demo