Non-determinism is fine at the edges. The system of record stays deterministic.

Anyone running AI agents in 2026 is hitting the same wall. Their procedures live inside system prompts. Their outputs are unstructured text. Two "identical" agents drift apart over weeks. Non-determinism is marketed as a feature when it should be a configuration option. And when something goes wrong, the debugging surface is a wall of narrative text.

The market is actively complaining about this — but nobody's selling the answer. The answer isn't a better agent platform. It's the layer underneath one.

Why a procedure server

Agents are runners. They're good at perception, conversation, judgment in the small. They are not the right place to store the work itself. The work — the ordered steps, the structured outputs, the rules — needs a home that isn't a prompt.

Hque is that home. You define procedures in a UI. Your agents — wherever they live, on whatever platform, in whatever language — call Hque to ask "what's next?" and report back what they did. The procedure is the source of truth. The agent is the runner. Hque is the record.

Once the work is structured this way, several things become free that used to be impossible: the same trigger produces the same procedure run every time. The output of step three can be queried, not parsed. A non-technical operator can edit the procedure without touching code. You can change your agent platform tomorrow without losing six months of operational history. And — almost as a side effect — you get a tamper-evident audit trail of everything that happened.

Why now

Three currents are converging in 2025–2026 that make this the right moment.

Agent platforms are proliferating, not consolidating. Zapier, n8n, Make, ElevenLabs, OpenAI, Anthropic, and a long tail of vertical AI SaaS. Every business is about to have agents running everywhere. The interesting bet is not which platform wins — it's the layer that survives all of them.

MCP is becoming the assumed interface. The Model Context Protocol is becoming the de facto standard for agent-tool integration. Building MCP-native from day one means automatic compatibility with whatever the next agent framework looks like.

The chaos is already biting. Operators with three-plus agents deployed have hit the prompt-soup, no-source-of-truth, can't-prove-what-happened problem already. They're not asking "should I worry about this?" They're asking "who's selling the fix?"

Who we are

Hque is built by Wain Digital, a small development house based on the Sunshine Coast in Queensland, Australia. Bryan Wain — founder, has been building enterprise software for twenty plus years and the platform that became Hque for four years. Orignally designed for humans to follow procedures, the thesis sharpened in late 2025 when the agent ecosystem made the AI Agent "system of record" gap impossible to miss.

The platform itself runs on .NET 9 / Aspire / Blazor WASM / MudBlazor / Azure. The audit trail is backed by Azure SQL Ledger. The in-product AI assistant — Que — uses the Anthropic Claude API and consumes the same MCP server that external agents do.

What we're betting on

→ That federation, not consolidation, wins. That the agent layer will churn for years, and the customer that bet on a single vendor will pay for that bet in migration costs. That the stable layer beneath the volatile ecosystem — the place procedures live and runs are recorded — is the right place to build a defensible business.

→ That vertical packs authored by domain experts will be more useful, and more defensible, than generic templates. That a dental practice doesn't want to invent its infection control procedure from scratch — but it also doesn't want the procedure trapped inside someone else's platform. That a channel partner with a network and a content asset is a better growth engine than another paid acquisition channel.

→ The audit trail — the cryptographic, tamper-evident record of what actually happened, isn't a feature most customers will ask for. It's the feature a small but meaningful minority will desperately need, and impossible to retrofit if you didn't build it from day one. So we built it from day one.