Frequently Asked Questions
GeraWitness — one inbox for every AI
What is GeraWitness?
GeraWitness is human-in-the-loop safety for AI agents. It intercepts agent-initiated transactions above a threshold you set (by amount, category, or counterparty) and routes them to one review inbox — web, mobile, Slack, or email. You approve, modify, or deny before the agent commits. Tamper-proof audit logs and optional agent-error insurance included.
Which AI agents does it support?
Any agent that implements the Model Context Protocol (MCP), ChatGPT Actions, or the GeraNexus agentic commerce protocol. Specifically supported at launch: Claude, ChatGPT with Actions, Gemini, Perplexity Pro, and all agent-enabled Gera Systems products. Custom agents can integrate via a single HTTP webhook.
How are thresholds configured?
Per agent: "GeraEats agent can spend up to £50 without approval, anything above needs a human signoff." Per category: "Any healthcare booking requires explicit approval regardless of price." Per counterparty: "Auto-approve transactions with merchants on the trusted list." Per time of day and per location are also supported. Thresholds can be updated from the web dashboard or API.
Is there agent-error insurance?
Yes, optional. Through GeraSure, a £4.99/month add-on insures agent-initiated transactions up to £1,000 per incident against agent error (wrong item, wrong quantity, wrong recipient, misinterpreted request). Claims are filed from the GeraWitness dashboard and processed within 48 hours.
Are the audit logs legally admissible?
GeraWitness logs are cryptographically signed and anchored to a public merkle tree daily. In jurisdictions that accept hash-anchored logs as evidence (EU, UK, US federal courts, most of the Commonwealth) the logs are admissible. The full audit trail includes: original prompt, agent interpretation, proposed action, review decision, final action, and outcome.