Automatic Video Face Anonymization MCP, Reversible Restore — NoonAI DIS

Core job-flow tools are clearly separated (upload vs submit vs poll status vs fetch result vs download file), and the one-shot encrypt_image/encrypt_video vs manual submit_encrypt_job overlap is explicitly explained in descriptions. Billing tools (get_billing_status, get_pricing_policy, get_usage_summary, purchase_credits vs purchase_credits_with_card) are closely related but distinguishable by their descriptions.

Almost all names follow a verb_noun snake_case pattern (upload_file, get_job_status, purchase_credits, register_billing_payment_method) with only minor variation like download_job_result adding a second noun. The mix of 'encrypt_*' one-shots and 'get_*_job_*' job tools is readable, though the vocabulary ('encrypt' vs 'job') is slightly loose.

15 tools is at the upper edge of the ideal range but each cluster (upload, job lifecycle, one-shot encrypt, billing) earns its place. The two upload paths and two purchase paths are justified by real client constraints (local file vs base64, stored method vs raw card).

Full lifecycle is covered: upload (path or base64), submit, poll, fetch metadata, download result, plus one-shot encrypt_image/encrypt_video and a complete billing subsystem. Minor gaps remain — no job cancellation, no asset/job deletion, and no explicit multi-job listing beyond get_usage_summary's ULID list.