Developer tools: data handling
The research and sandbox tools need only country, capability, business event, synthetic scenario, fixture ID or reason code. Do not send names, identity numbers, bank account numbers, documents, passwords or API keys. Remote MCP and HTTP requests are processed on Gridzen servers. The developer service has no user database and does not retain tool request bodies or run provider checks. Network infrastructure may process connection metadata such as IP addresses. Local stdio runs offline after dependencies are installed. Third-party clients, package registries and installers have their own data policies. Contact open@gridzen.ai about privacy or deletion questions. Remote MCP stores aggregate call counts without personal data: UTC date, tool name, fixed task/capability/country/stage/source/error categories and counts. It does not store request contents, identifiers or IP addresses in this separate counter database. Rows aged 90 days are removed on the next write; local stdio is not counted. Counts are not unique users. Updated 2026-10-09.
Task evidence and optional feedback
Remote APIs aggregate only fixed task, capability, country, stage, source and error categories for 90 days. No raw prompts, personal data, credentials or user tracking are stored. Offline Skills do not upload automatically. Optional summaries require explicit consent and remain self-reported. Receipt hashes expire after 30 days. Technical requests do not establish unique customers or completed business tasks.