Enterprise

What the company is asking for

The same twenty questions arrive every week: what the policy says about a contractor working remotely, whether an offer is complete, which invoices exceed the limit. The answers exist — in a policy nobody rereads, a system nobody outside the team can query, and the memory of two long-serving colleagues.

• Answer a policy question with the clause it comes from, not with an opinion.
• Compare supplier offers against the procurement policy and say what each one is missing.
• Check a batch of invoices against the approval limits and flag the exceptions.
• Prepare the file a manager needs before an approval, instead of after it.

How the work is done

An agent is a model, a set of approved skills, tools and a policy, applied to a scope.

Agents answer from your own regulations, documents and records, with the source attached.

Agents reach your systems through approved connectors — APIs, databases, applications — each with its own permissions.

Changes wait for human approval before they run.

Customer-controlled inference and execution boundaries.

Every one of those systems runs somewhere the IT team is accountable for. That estate is also where a company can satisfy itself about the controls before a single HR document is loaded.

Read-only execution policy is the default on every enrolled host; the full policy is an explicit, per-host choice.

Every answer shows the exact commands that ran, their exit codes and output, per host.

A host joins the fleet with one install command and an enrollment token; revoking the host unpairs it.

Bring your own model — OpenAI-compatible endpoints, Ollama, Anthropic, or Cloudflare AI Gateway / Workers AI — with keys envelope-encrypted at rest.

The whole stack self-hosts with one command; an air-gap bundle ships every image.