Docs / ai / Available Tool Categories

Available Tool Categories

guide

The nine capability categories you declare on a registered agent (and, advisorily, on useAgent), each granting a group of related tools.

Declare these in capabilities when registering an agent via casino_create_agent. (For persona-backed apps the registered definition is authoritative — categories passed to useAgent are advisory and cannot widen the persona's grant.)

- data_access — query_mongodb, insert_documents, update_documents, aggregate_mongodb, count_documents, list_collections, query_data_source, and more - external_integration — web_search_deep, google_ads, google_analytics (GA4), google_drive, twitter, reddit, youtube, strava, kaggle, call_integration, send_app_sms, register_app_phone - site_operations — build_project, write_project_file, read_project_file, get_site_files, publish_version, edit_site, import_github_repo - code_execution — python_exec, daytona_run_code, materialize_in_sandbox, claude_code_exec - domain_task — finance/health/legal/tax domain tools, fetch_daily_brief, text_extract_pattern - workspace_management — save_document, list_documents, search_documents, save_report, save_visualization, remember, search_memory - agent_management — spawn_subagent, create_subagent, list_subagents, list_subagent_runs - specialist_agents — data_science, legal_review, app_building, tax_analysis, accounting, route_to_agent - automation_management — create_automation, list_automations, pause/resume/delete/run automations

Most built-app agents only need data_access. Note web_search (basic) is always available and is not gated — only web_search_deep requires external_integration. Run casino_list_tools (MCP) for the live, authoritative tool list per category.

See also

targeting-custom-agents · tool-access-control · agent-catalog

Source: README.md#available-tool-categories · Also available in llms-full.txt and registry.json.