Becoming Nex0 Oy — and giving every agent a solver

Two announcements this month, one administrative and one architectural.

The administrative one: as of January 2026, this business operates as Nex0 Oy, incorporated in Helsinki. Vera began as a sole-trader operation in 2025 and grew on customer revenue; incorporation reflects where the company is heading — more customers, a larger team, and the contractual machinery that serious logistics organizations rightly expect from a vendor (DPAs, liability frameworks, the unglamorous load-bearing paperwork).

The architectural one is bigger. Vera 1.2 ships an MCP server.

The Model Context Protocol is emerging as the standard way for AI agents to call tools. With 1.2, Vera's reasoning engine — rule parsing, plan evaluation with unsat cores, memory queries, full verified decisions — is exposed as MCP tools that any compliant agent framework can call.

We did this because of a conversation that kept repeating. Companies are building their own agents; they are excited and slightly terrified, because their agents have the same problem every pure-LLM agent has: nobody can prove the agent won't violate a constraint. Our answer used to be "replace your agent with Vera," which is a hard sell to a team that just built one.

The new answer is better: keep your agent, give it a solver. Your orchestrator handles the conversation and the workflow. Before it commits to anything operational, it calls vera_evaluate_plan. If the plan violates a rule, it gets back the exact constraint that failed and can fix precisely that. Your agent becomes provable at the only moment that matters — the moment of action.

Autonomy ceilings and audit logging apply to MCP callers exactly as they do to our own console. A third-party agent cannot exceed the autonomy level the customer's operators configured. The trust model doesn't bend for convenience.

One company, one engine, two ways to consume it: as a product for operations teams, as a component for engineering teams. 2026 is about proving both at scale.