What we do
We take one operational workflow, such as customer cases, refunds, payouts, compliance checks, treasury operations or access requests, and put an AI agent to work on it. The first workflow goes live in 2–4 weeks from kickoff. Then we run the deployment, keep the rules current and add the next workflow. Customers keep their tools, their model vendor and the rules they already have.
The work starts with a readiness check: four questions, then a one-page plan by email naming the workflow an agent can take over first, what blocks it today and what it takes to go live.
How we build it
- Denied unless allowed. Every action is checked against the customer’s rules and approval steps before it runs. Anything not explicitly allowed is denied, and the attempt is recorded.
- The agent never holds credentials. Passwords, API keys and wallet keys stay in FeirOS, which uses them on the agent’s behalf once an action is approved.
- A record anyone can check. Each action is recorded with who approved it and signed, so any change shows. An auditor can verify the record on their own machine, without us.
- Limits stated up front. Each agent has a spend limit with a small, stated overshoot, and one switch stops it.
- In the customer’s environment. FeirOS runs on infrastructure the customer controls. We have no standing access; access during a deployment is granted, revocable and on the same record.
Who we work with
Heads of Operations and COOs at EU crypto and fintech companies of roughly 20 to 200 people: licensed crypto-asset service providers, exchanges, stablecoin issuers, crypto payment and on-ramp providers, payment and e-money institutions, neobanks, and lending and BNPL firms. Banks, insurers and other supervised firms are next. Read how it works for crypto operations, payment operations and other regulated operations.
FeirAI at a glance
| What we do | Deploy and run AI agents in the operations of regulated EU crypto and fintech firms |
|---|---|
| Product | FeirOS, the control layer between AI agents and the systems they act on |
| Open source | Vultrino, which holds credentials, and Averin, which keeps the record |
| Where it runs | In the customer’s environment, on infrastructure the customer controls |
| In production | Customer operations at an EU crypto company, not named publicly |
| Works with | Agents built on LangChain, Hermes or custom code, in Python or TypeScript, with the model vendor the customer already uses |
| Regulation | Built for firms under DORA, MiCA and the EU AI Act. FeirOS enforces rules and keeps a verifiable record; it does not replace a compliance team |
| Certification | No SOC 2 or ISO 27001 yet; on the roadmap |
| Markets | The EU and EEA, in English |
| Contact | contact@feir.ai |
Check our work yourself
The parts of FeirOS that hold keys and keep records are open source, so your security team can read them: Vultrino and Averin. The record verifier checks any FeirOS record, and the same tool runs offline. The live demo walks one action through the operator console. The guides explain the reasoning behind each control, with sources.
What you keep, whatever happens to us
The software runs in your environment. The parts that hold keys and records are open source. The rules are yours, and every record verifies without us. If FeirAI changes or goes away, your deployment keeps running.
Leadership
FeirAI was founded by Dominik Zachar, previously CTO of Cypher Enterprises and solution architect of the Summon Platform, the first DAO platform on Cardano. That work was community, treasury and governance infrastructure for votes of up to about 100,000 concurrent participants, where every action moved real funds. The same questions carry over to AI agents: who may act, under which rules, with whose approval, within what limits, and how anyone can check afterwards what happened.
Start here: the free readiness check takes about two minutes. Or email contact@feir.ai.