What is an Agent Data Residency?

Data residency becomes an agent-specific problem because agents don't just store data at rest — they move it through multiple hops during a task: into an LLM provider's inference endpoint, into a sandbox container, into logs and telemetry, and out again to whatever tool it calls next. A residency guarantee that only covers the database and ignores where the model inference or the sandbox runtime physically runs is incomplete, and this is exactly the gap regulators look for under GDPR's international-transfer rules or sector-specific mandates like data-localization laws. Satisfying it in practice means knowing, and being able to prove, which cloud region every hop in the agent's execution path runs in, restricting which LLM providers and MCP tool endpoints an agent can call when residency is required, and keeping that restriction enforced at the infrastructure level rather than as a policy nobody checks.