Introducing Tool Calling in Copilot

The Auralius Copilot panel now supports tool calling. Instead of reading dashboards and manually triggering actions, you can describe what you want in plain English and the Copilot will execute it — with a full audit trail and a step-up gate for anything high-risk.

Available tools

• debug_agent — inspect the last N turns of a live agent, including model chosen, tokens, latency, and RAG sources.
• query_knowledge_graph — run a hybrid vector + BM25 search against any of your knowledge bases and see ranked chunks with scores.
• deploy_channel_route — wire a phone number or channel to an agent with a single command.
• rotate_secret_keys — rotate API keys. This is a HIGH-risk tool and requires a TOTP step-up.

The human-in-loop gate

Every tool tagged HIGH or CRITICAL opens a confirmation modal that asks for your current TOTP code before executing. The execution — whether approved or denied — is written to an immutable SHA-256 hash chain. You can verify the entire chain at any time via GET /mcp/audit/verify.

Credit model

Each tool execution consumes 1 Copilot credit. Credits are included in your plan and can be topped up in increments of 100 or 300 on the Billing page.