Metasvit is a blockchain and AI-agent engineering studio in The Netherlands. Smart contracts, autonomous agents, DeFi routing and GameFi — shipped since 2021.
Each is an area we have shipped repeatedly and can show you the repositories for — scoped, priced and delivered as a stand-alone engagement.
An agent is only useful if you can trust it unattended. Ours ship with a risk layer that can halt the loop, structured logs you can replay decision by decision, and a kill switch a human can reach in one step.
Multi-year engagements rather than one-off builds — infrastructure, games and trading systems that reached production and stayed there.
Wallet, KYC gateway and game client for an XRP Ledger ecosystem.
A full product surface on the XRP Ledger: a browser wallet, a KYC gateway for compliant onboarding, and a Unity game client with wallet authentication — plus the container and Kubernetes infrastructure the services run on.
GameFi platform: WebGL client, live leaderboard and NFT tooling.
A play-to-earn title end to end — the WebGL game client, a .NET API behind it, a live leaderboard, and the NFT tooling for generating and minting the collection that backs the in-game economy.
A DEX front end with its own smart order routing service.
A swap interface backed by a routing service we built and ran ourselves, so quote quality and uptime were never dependent on a third-party endpoint.
Agent runtimes for on-chain execution, with human escalation.
Our current line of work: agents that execute on-chain tasks unattended, paired with notification services that escalate to a human for the decisions that still need one.
We read your contracts, your threat model and your roadmap, then write down what we think you are actually building. If we disagree with the plan, you hear it here — not after the invoice.
Two-week increments against a shared board. Contracts land with tests and a gas report; agents land with a replayable log. You can run what we have at the end of any sprint.
External audit, testnet soak, and the remediation pass afterwards. We write the upgrade and rollback path before mainnet, not after the first incident.
Deploy, monitor and stay on call. Agents get alerting and a kill switch a human can reach in one step, because unattended does not mean unsupervised.