13 new tools across read/safe-write/destructive tiers. Total: 28 tools (11 reads / 12 safe-writes / 5 destructive).
Tier 1 reads (4)
- adguard_check_host - test what AGH would do with a hostname
- adguard_get_blocked_services - global blocked services + weekly schedule
- adguard_get_dns_config - upstream DNS, cache, blocking mode
- adguard_get_safesearch_settings - SafeSearch state + per-engine flags
Tier 2 safe-writes (6)
- adguard_refresh_filter_lists
- adguard_add_client
- adguard_update_client
- adguard_set_blocked_services (PUT, accepts HH:MM or ms)
- adguard_toggle_safesearch (PUT)
- adguard_toggle_safebrowsing
Tier 3 destructive (3)
- adguard_delete_client
- adguard_clear_query_log
- adguard_reset_stats
Highlights
- 92/92 tests pass (+32 since v0.1.1)
- Both mcp-server.ts and index.ts wire through buildAllTools() (single source of truth)
- add_client uses explicit allowlist; gate flags cannot leak into AGH body
- toggle_safesearch forces per-engine false when disabled
Install: `npm install @solomonneas/adguard-mcp`