Privacy
Aurora MCP is a thin broker. It forwards your queries to existing Consilio APIs and returns the results to your AI assistant. It does not accumulate data of its own.
What the MCP server sees
Section titled “What the MCP server sees”- The identity headers injected by the gateway (your Consilio sub, email, groups, authentication assurance level).
- The tool name being invoked and the arguments the client chose.
- The upstream API response, which is returned to the client.
What the MCP server keeps
Section titled “What the MCP server keeps”- Operational metrics — tool name, response latency, response size, status, caller sub — retained for a limited window for performance and abuse detection.
- Error traces — redacted of argument content, used for debugging outages.
What the MCP server never persists
Section titled “What the MCP server never persists”- Prompt content from the client.
- Tool outputs (document contents, matter data).
- Access tokens (the server never sees them; they stop at the gateway).
Data residency
Section titled “Data residency”Aurora MCP runs in the same infrastructure footprint as the upstream Consilio APIs it calls. No data leaves that footprint as a side-effect of using this connector.
Your contract with your AI provider
Section titled “Your contract with your AI provider”When you use an AI assistant, your provider’s privacy terms govern what they do with your prompts and the assistant’s responses. Aurora MCP does not change those terms — it only provides data your assistant can cite.