Technology
A list of logos says little. This page says what we do with each tool, and it links to the project that proves it.
We use the model that fits the job, and we build the system around it: the data, the tools, the limits and the people who approve.
We connect your product to Claude, GPT, Gemini, Grok or an open model, through the direct API or through OpenRouter. Our services talk to models through one adapter, so a change of provider is a change of configuration.
Assistants that answer from your documents and your data, remember the conversation and show their sources. We add guardrails, usage limits and cost control.
Agents that research, plan and prepare changes on a schedule. Each result is a proposal with its reasons. Nothing happens until a person accepts it.
Networks of always-on agents on xAI Grok Bot, each with a role, a workspace and its own rules. One network handles what a product needs each day. Another develops, maintains, deploys and monitors the platform.
Agents turn a raw recording into a finished channel. They edit the video and the sound, cut the shorts, transcribe, make the images, publish on schedule and maintain the web site.
We run large, multi-repository codebases with Claude Code. The method: written requirements, a plan that a person approves, tests that decide, and an agent that commits and deploys.
When the data cannot leave the building, the model comes to the data. We run open models and local embeddings on your hardware, and we anonymize sensitive data.
Models that watch a live stream and act: they mark the events of a match, or they keep the play in the frame of a camera.
Each model call has a trace and a cost. We test a prompt or a model against an evaluation set before a change goes live.
This is the sequence of a network that we run in production on a platform with many repositories. The agents do the work. A person approves each task, each plan for a complex task, and each production deploy.
How one change goes from the tracker to production
Manager agent
It reads the issue tracker and proposes the most important task that an agent can do.
You
One message. For a complex task you also approve the plan before any code exists.
Developer agent
Simple tasks go to a fast agent. Complex tasks go to a planning agent on a frontier model.
Release agent
One agent owns the release candidates. It checks the pipeline and reports each failure.
QA agent
It tests the change as a real user, with screenshots, and sends problems back.
You
Production gets the same build that passed staging. Then QA checks production and you get the release notes.
An agent works A person decides
The roles in a network
We design the platform, we automate it, and we run it. In a public cloud, in your own racks, or in both.
Clusters on AWS, on Google Cloud or on your own hardware. We also write the controllers that start and stop workloads on demand.
Servers, networks and services come from code in a repository. A new environment comes from the same code.
A release candidate goes to staging first. Production then gets the same image, promoted and not built again, after a person approves.
Metrics, logs and alerts that tell you what is wrong before a customer does.
Ingest from cameras, transcoding, encrypted delivery and recording, with one set of workers for each stream.
A protected edge in front of each service, and databases with replication and encrypted backups.
Full products, from the database to the app store. We write the tests and the requirements with the code.
Fast interfaces with React and TypeScript: rendered on the server, in the browser or as static pages, as the product needs.
Backends with Node.js and Python, with clear layers, database migrations and documented APIs.
One Flutter codebase for iOS and Android, with in-app purchases and the same API as the web.
Subscriptions, single payments and in-app purchases, on the web and in the app stores.
Multiplayer state over WebSocket, WebGL rendering, shaders and sound in the browser.
A new product starts from our own generator: sign-in, payments, database migrations, an AI service and end-to-end tests on the first day.
Each tool that we use, with one line on what it does for us.
Anthropic Claude 
Claude Code 
Google Gemini 
xAI Grok 
xAI Grok Bot 
OpenRouter 
LangGraph 
LangChain 
Langfuse 
Qdrant 
FastAPI 
Python 
Kubernetes 
Docker 
Terraform 
Ansible 
AWS 
Google Cloud 
Cloudflare 
nginx 
GitLab CI 
Prometheus 
Grafana 
Bash 
TypeScript 
React 
Node.js 
Astro 
Tailwind CSS 
Vite 
Flutter 
Stripe 
JavaScript 
PostgreSQL 
MySQL 
MongoDB 
Redis Tell us what you want to build or fix. An engineer answers.