Adrails

The analytics/reporting surface is heavily crowded: reporting_overview, performance_period, performance_diagnose, metrics_daily, portfolio_review, findings_review, changes_review, actions_recent, business_review, budget_recommend, planning_calculate and metric_compare all read or explain performance, and several describe themselves by which other tool to read first rather than by a distinct capability. Similarly site_brief vs brand_read_site and store_orders vs store_revenue vs metrics_daily require the agent to absorb long caveats to pick correctly. The descriptions do the disambiguating work, but the boundaries remain blurry for a first-pass selection.

Names follow a predictable domain_prefixed pattern (store_*, creative_*, campaigns_*, integrations_*, workflows_*, knowledge_*, performance_*, google_*), which makes the set scannable. Minor deviations exist (accounts_list/actions_recent put the verb second while others like brand_read_site put it third, and campaign_drafts vs campaigns_setup differ in number), but the overall convention holds.

45 tools is well beyond a comfortable surface and reflects several loosely related domains (ad platforms, commerce, SaaS billing, workflows, knowledge, integrations) bundled into one server. Many tools are narrow read-only slices (editor_settings_reference, integrations_catalog, integrations_detail, targeting_options) that pad the count without earning distinct places.

Several tools explicitly point at companions that are absent from this surface — campaign_drafts references campaigns_publish_draft, launch_brief references campaigns_launch, knowledge_propose references knowledge_manage, workflows_list references workflows_manage, and actions_recent implies an undo tool that is not listed. The result is a surface rich in reading and preparation but with dangling handoffs at the points where an agent would actually need to act.