Archive Technologies Rate Limits

Two limits apply to an Archive workspace at the same time, and whichever is reached first binds: a per-plan CREDIT BUDGET that refills every second, and a flat ceiling of 5 requests per second. Both are per workspace and never shared — one token covering several workspaces gets a full budget in each, including on agency plans. One credit is one millisecond of backend compute, so cost follows the work a request causes rather than the size of the response. Archive's hosted MCP server draws on the same budget: a tool call maps to a query and costs what that query costs. Every response carries the IETF ratelimit / ratelimit-policy headers, so a client can read its plan, its burst allowance and its remaining credits off any response instead of discovering the ceiling with a 429.