What the control room shows

This is the private operations view of your spend firewall. Paste the API key delivered after checkout to load only that key's isolated workspace. The counters and activity refresh every 15 seconds.

Live activity lists every evaluation with its amount, decision badge, merchant, and the exact rule that produced the decision. Approvals is the human-in-the-loop queue: transactions that crossed your approval threshold wait here until you approve or deny them. Rules is the policy editor with seven rule types — per-transaction caps, daily totals, velocity limits, merchant block and allow lists, category limits, and approval thresholds — each with its own parameters, action, and priority.

The Test API tab sends a real request to the same /v1/transactions/evaluate endpoint your agents call, so you can watch a $750 purchase from an unknown merchant get blocked, then see the decision land in the feed and the audit log. Prefer a guided demo? The public playground runs the same endpoint with preset scenarios, and the framework integrations show the five-line client for LangChain, CrewAI, the OpenAI Agents SDK, and the Vercel AI SDK.

Every decision shown here is written to the workspace audit log, and the engine behind it passes the public eval suite. Hosting starts at $99/month; the MIT-licensed core can also be self-hosted.