Field notes on building
and governing AI agents.
Architectural arguments, operational diagnostics, and the editorial through-line of the Alquimia Agentic Platform. Long-form, board-readable, engineer-honest.
Attacking Your Own Agent: MiDojo, Roast Me, and When Each One Applies
Why an agent testing tool from Red Hat and the library we develop inside Gaussia answer different questions, and how to tell which one a deployment needs.
Proof Before Action: How Alquimia Enforces Agent Authorization with Zero Trust and MCP
Alquimia combines Zero Trust policy, signed channel identity, MCP-owned approval, and UCP checkout semantics so one informed confirmation authorizes one exact tool execution.
Zero Trust for AI Agents: How We're Rethinking Security at Alquimia
Traditional zero trust was built for servers and networks, not systems that plan, reason, and act on their own. Here's how we're extending it — from non-human identity for every agent to designing as if the attacker is already inside — to secure autonomous AI at Alquimia.
Ledger: knowledge our agents can be audited on
Every layer of an agent stack has to be governable, and knowledge is the weakest one today. We are building Ledger, a module that compiles curated sources into versioned, OKF-conformant knowledge bundles that agents consume through MCP — so what an agent knows becomes an artifact you can check out, diff, and audit.
Running AI agents on Red Hat OpenShift AI: lessons from a sovereign deployment
GPU economics, hardware observability, model explainability — three runtime-layer choices that decide what governance is feasible above the agent layer.
Why governing AI agents end-to-end is now a board-level concern
When AI agents make decisions a person used to be accountable for, governance reaches the boardroom. The six properties every audit committee should ask about.
From a notebook to a fleet: why AI agents in production need a platform layer
The first AI agent is not the hard one. The fourth is. When one agent becomes a fleet, the platform underneath them is what makes it operable.
Run production-grade agents you govern end-to-end.
We work with enterprise teams building AI agents on their own infrastructure. A short call is enough to see if Agentic Platform is the right fit for your case.
Get in touch