Getting Started

Open the workflow tab and launch your industry's scenario. The published Studio workflow runs end to end against the Callvu Sandbox tenant. Before a single screen renders, a pre-flight gate checks the session against the vertical's bound regulation, six evaluator agents score the request across completion risk, compliance evidence, policy interpretation, data trust, audit readiness, and exception patterns, and a deterministic orchestrator issues the directive.

In Banking, that is an account address change bound to GLBA. A verified session reaches the Address Change MicroApp, which collects the new address and the customer's consent, and any compliance edge case escalates to a human agent instead of executing. You can launch the workflow in a web or mobile view, and every run is observable in the Analytics tab.

Drive the same workflow conversationally. A run started from the chat is evaluated before it starts by the same six agents and the same policy as the workflow tab, and produces the same Decision Trace and Audit Trace.

Stick to your vertical's scenario. In Banking, ask to change your address on record. Any username works; type skip as the password. Choose any US state except New York and answer Yes for a clean approval. Choose New York to trigger an escalation: a rule in the workflow sends New York changes to a human reviewer. Answer No to a consent question to see a declined change.

The Analytics tab is the compliance surface. It has four subtabs. Forms and Workflows show submission and execution volume over the selected time window. Agents lists every agent session; click one to open the full per-session decision drilldown. Audit Trace lists every workflow execution and its governance record. Two artifacts come out of here, and both download as Callvu-branded PDFs.

Every invocation produces a Decision Trace: the reasoning behind a single AI action. Open any session under Agents to see the invocation context, each of the six evaluator agents with its recommendation, risk score, and confidence, and the orchestrator's final directive. Download the full set from the Agents subtab as a Callvu-branded PDF that includes the full JSON record.

Every workflow execution produces an Audit Trace: the chronological governance record of what happened. Open any execution under Audit Trace to see each checkpoint, the policy it enforced, and whether it escalated, timestamped and in order. Download the executions for the selected time period from the Audit Trace subtab as a Callvu-branded PDF, each execution boxed and citing the vertical's regulation at every step.

Point your own AI backend at the Callvu Sandbox tenant with a pre-issued API key, no Studio login required. Your AI is now constrained by the same runtime: the same agents evaluate, the same policies enforce, and the same Decision Trace is produced for every action it takes. Full instructions and the key are inside the Bring your own AI tab after you unlock.

Your AI talks to the Callvu Chatbot Connector with two calls: one to find the workflow, one to start it. The tab gives you the exact requests, and a one-file starter chatbot in Node or Python with your vertical's workflow already filled in. Add your Callvu key and an Anthropic key, run it, and chat.

Want to see it first? “Run it live with Claude” in the same tab has Claude start your vertical's workflow in front of you and shows every call it makes and the evaluators' decision. Five runs per visitor per day.

To see a session that fails the check, pass "cond": "0" in initialData, or press “Run a failed check”. The form never opens and a Compliance Governance step escalates the session to a human agent; the Audit Trace for the run shows it. Only your own AI can do this; the link and the chat cannot.