Run the best agents. Surrender nothing.
Developers should be able to use the best AI coding agents without giving up their code, their choice of models, or their continent's ability to build this infrastructure itself. That is the whole mission: frontier agentic coding on your own terms, and a Europe that builds the AI layer instead of only consuming it.
Five things we won't trade away.
Local-first is an ethic, not just an architecture.
Your code stays on your machine; sessions and threads live locally; agent requests go straight to your model provider with your keys. We built it that way because it is better engineering, and because software that respects you should ask for inspection, not trust.
Say only what is true.
We write readiness, not certified, until the auditors have signed. No fake customer logos, no invented metrics, no praise we did not earn. A claim we cannot back is a bug, and we treat it like one.
The harness, not the cage.
Radian includes its built-in agent and runs 19 external coding agents today - Codex, Claude Code, Cursor, upstream Pi, OpenCode, GitHub Copilot, Grok Build, Cline, and the rest of the public roster - each on its own install, auth, and model path with your own subscriptions and keys. If a better agent appears tomorrow, Radian should become the best place to run it. Lock-in is a business model, not a feature.
Engineered in Europe, built for everywhere.
We are a European company and glad of it: GDPR is our starting point, not a retrofit, and Arc serves open models from European-owned infrastructure. But we compete on craft, not geography; the tool has to win on its own.
Small team, high craft.
Few people, senior work, high trust. We would rather ship one calm, finished surface than ten busy ones. When in doubt, remove. Small is not a stage we are passing through; it is how we intend to work.