Built-in Tools Reference¶

This page documents all built-in tool categories available to chat in the ASK Assistant Service (ask-assistant-service), including which model each tool uses, how model references are wired, and how each tool is enabled or disabled.

How Tools Reference Models¶

All tools follow a single wiring chain:

Tool code
  → get_tools_settings().<tool>_llm property
    → settings.llm_model_name (env var override)
      → Foundry API (AI71_FOUNDRY_ENDPOINT)
        → models.yml routing → actual provider (Anthropic / OpenAI / etc.)

Key files:

File	Purpose
`src/tooling/config/settings.py`	Central `ToolsSettings` class — all model names as env-var-backed fields with named properties per tool
`src/tooling/core/mcp_servers/servers/*.py`	One file per tool server — declares `always_available`, `requires_auth`, and `ServerRequirement` (env var / app config / server dependency gates)
`config/default.toml`	Dynaconf defaults for tools that use nested config (data_analysis, meta_agent, image_analysis)
`src/model_provider/models/models.yml`	Foundry model registry — maps model string IDs to upstream providers

Model name fields in ToolsSettings (all overridable via env var):

Field	Env Var	Default
`llm_model_name`	`LLM_MODEL_NAME`	`azure/gpt-5-chat`
`small_llm_model_name`	`SMALL_LLM_MODEL_NAME`	`azure/gpt-5-chat`
`code_generator_model_name`	`CODE_GENERATOR_MODEL_NAME`	`azure/gpt-5-chat`
`docx_prepare_context_model_name`	`DOCX_PREPARE_CONTEXT_MODEL_NAME`	`qwen/qwen3-5-122b-a10b`
`data_visualization_llm`	`DATA_VISUALIZATION_LLM`	`anthropic/claude-3-7-sonnet`
`foundry_tool_embedding_model_name`	`FOUNDRY_TOOL_EMBEDDING_MODEL_NAME`	`BAAI/bge-large-en-v1.5`

Named properties on ToolsSettings map each tool use-case to the right field:

@property
def web_search_llm(self)            → self.llm_model_name
def summarization_llm(self)         → self.small_llm_model_name
def translation_llm(self)           → self.llm_model_name
def code_generator_llm(self)        → self.code_generator_model_name
def docx_prepare_context_llm(self)  → self.docx_prepare_context_model_name
def extraction_llm(self)            → self.llm_model_name
def email_query_extraction_llm(self)→ self.llm_model_name
def people_extraction_llm(self)     → self.llm_model_name

To swap the model for any tool, set the corresponding env var — no code changes needed.

Enable / Disable Legend¶

Status	Meaning
✅ Always on	`always_available=true` on the MCP server — active for all agents with no configuration
🟡 Conditional	Requires an env var, app config block, or per-agent activation
🔐 OAuth required	`requires_auth=true` — user must complete an OAuth flow; plus provider env vars must be set

Complete Tool Reference¶

Category	Sheet Feature Name	Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
Email (Gmail)	—	ThreadSearchTool, FetchEmailsTool, FetchRelevantEmailsContentTool, SelectRelevantEmailsTool, ComposeAndExtractEmailTool, CreateUpdateDraftTool, SendGmailEmailTool, FetchAttachmentContentTool, EmailConsistencyCheckTool, BreakdownThreadFetchingQueryTool	`email_query_extraction_llm`, `thread_fetching_breakdown_llm`, `relevant_content_extraction_llm`, `extraction_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; Google OAuth flow; requires `GOOGLE_GMAIL_CLIENT_ID` + `GOOGLE_GMAIL_CLIENT_SECRET`
Outlook Email	—	SearchOutlookEmailsTool, FetchOutlookEmailsTool, SendOutlookEmailTool, CreateOutlookDraftTool	`email_query_extraction_llm`, `thread_fetching_breakdown_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; requires `MICROSOFT_CLIENT_ID` + `MICROSOFT_CLIENT_SECRET` + `MICROSOFT_TENANT_ID`
Outlook Calendar	—	FetchOutlookEventsTool, CreateOutlookEventTool, UpdateOutlookCalendarEventTool, CancelOutlookCalendarEventTool	`extraction_llm`, `people_query_breakdown_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; same Microsoft OAuth credentials as Outlook Email
Calendar (Google)	—	CreateCalendarEventTool, UpdateCalendarEventTool, CancelCalendarEventTool, ListCalendarSeriesTool, GetAvailabilityTool, CalendarConsistencyCheckTool, BreakdownCalendarQueryTool, ListCalendarsTool, ComposeAndExtractCalendarEventTool	`extraction_llm`, `thread_fetching_breakdown_llm`, `people_query_breakdown_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; requires `GOOGLE_CALENDAR_CLIENT_ID` + `GOOGLE_CALENDAR_CLIENT_SECRET`
Google Docs	—	CreateGoogleDocTool, ReadGoogleDocTool, UpdateGoogleDocTool, DeleteGoogleDocTool, CopyGoogleDocTool, ListGoogleDocsTool	— (Google API calls)	none	—	🔐 OAuth required — `requires_auth=true`; requires `GOOGLE_DOCS_CLIENT_ID` + `GOOGLE_DOCS_CLIENT_SECRET`
People	—	SearchPeopleTool, EnrichPeopleTool, BreakdownPeopleQueryTool, Google Contacts enricher, Directory searcher	`people_extraction_llm`, `people_search_extraction_llm`, `people_query_breakdown_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; activated together with email/calendar/docs servers
Web Search	Orchestrator Model	WebSearchTool, WebCrawlTool, ArxivSearchTool, ArxivUrlTool	`web_search_llm`	`azure/gpt-5-chat`	`llm_model_name`	🟡 Conditional — requires env var `WEB_SEARCH_TAVILY_API_KEY` (`strict=false`, degrades gracefully without it)
Code Execution	—	CodeInterpreterExecutionTool, XlsxSkillTool, Daytona runtime/artifacts/helpers	— (no LLM call)	none	—	🟡 Conditional — `always_available=false`; enabled per-agent; no hard env requirement
Daytona Minimal	—	DaytonaMinimalCodeExecutionTool	— (no LLM call)	none	—	🟡 Conditional — bundled with code_execution server; same activation rule
Code Generator	Orchestrator Model	CodeGeneratorTool	`code_generator_llm`	`azure/gpt-5-chat`	`code_generator_model_name`	🟡 Conditional — `always_available=false`; also required by data_analysis as a `SERVER_DEPENDENCY`
Data Analysis	Model used by various tools	DataAnalysisTool (column_mapper, data_profiler, query_validator sub-utils)	`app_config.tooling.data_analysis.*` (toml)	`azure/gpt-5-chat` (all 3 sub-models)	`app_config.tooling.data_analysis.{profiler,code_generator,column_mapper}_model`	🟡 Conditional — `always_available=false`; `SERVER_DEPENDENCY` on both `data_visualization` and `code_generation` servers
Data Visualization	Model used by various tools — Data visualization tool	DataVisualizationTool	`data_visualization_llm`	`anthropic/claude-3-7-sonnet`	`data_visualization_llm`	🟡 Conditional — `always_available=false`; also a `SERVER_DEPENDENCY` for data_analysis
Image Analysis	KMS models — OCR Model	ImageAnalysisTool	`app_config.image_analysis.model` (toml)	`Qwen/Qwen3-VL-8B-Instruct` (VLM)	`app_config.image_analysis` (VLM config block)	🟡 Conditional — disabled by default (`image_analysis=@null` in `default.toml`); must be set in env-specific toml; `RequirementType.APP_CONFIG`
Database Navigator	Orchestrator Model	DatabaseNavigatorTool	`llm_model_name` / `small_llm_model_name` (fallback chain)	`azure/gpt-5-chat`	`llm_model_name` → `small_llm_model_name`	🟡 Conditional — `always_available=false`; no hard env requirement; enabled per-agent
PP DB Connector	—	ListDatabaseConnectionsTool, ExecuteDatabaseQueryTool, GetDatabaseConnectionTool	— (no LLM call)	none	—	🟡 Conditional — `requires_auth=false`; toggled by org/agent configuration
Knowledge Base	KMS models — Embedding Model	RetrievalTool (KMS chunks, documents, retrieve)	— (vector retrieval)	`BAAI/bge-large-en-v1.5` (embedding)	`foundry_tool_embedding_model_name`	🟡 Conditional — requires `KMS_BASE_URL`; enabled when a knowledge base is attached to an agent
Knowledge Extraction	Orchestrator Model	DocumentKnowledgeExtractorTool	`extraction_llm`	`azure/gpt-5-chat`	`llm_model_name`	🟡 Conditional — requires `KMS_BASE_URL`; bundled with knowledge_base activation
Summarization	Orchestrator Model	SummarizationTool	`summarization_llm`	`azure/gpt-5-chat`	`small_llm_model_name`	🟡 Conditional — `always_available=false`; enabled per-agent
Translation	Orchestrator Model	TextToTextTranslationTool, DocumentToDocumentTranslationTool	`translation_llm`	`azure/gpt-5-chat`	`llm_model_name`	🟡 Conditional — requires env var `KMS_BASE_URL` (`RequirementType.ENV_VAR`)
Google Maps	—	GoogleGeocodingTool, GoogleReverseGeocodingTool, GoogleDirectionsTool, GooglePlacesCategorySearchTool, MapVisualizationTool	— (Google Maps API)	none	—	🟡 Conditional — `always_available=false`; requires env var `GOOGLE_MAPS_API_KEY` (`RequirementType.ENV_VAR`)
OSM	—	OSMGeocodingTool, OSMReverseGeocodingTool, OSMDirectionsTool, OSMCategorySearchTool, MapVisualizationTool	— (Nominatim API)	none	—	🟡 Conditional — requires env var `OSM_USER_AGENT` per Nominatim usage policy; no auth needed
Online Media	—	YouTubeTool, YouTubePlaylistTool	— (YouTube API)	none	—	🟡 Conditional — requires env var `OXYLABS_PROXY_ENDPOINT` (`RequirementType.ENV_VAR`) for proxy-based YouTube access
DOCX	Orchestrator Model	DocxPrepareContextTool, DocxApplyOperationTool, DocxMarkdownToWordTool	`docx_prepare_context_llm`, `code_generator_llm`	`qwen/qwen3-5-122b-a10b` (prepare) / `azure/gpt-5-chat` (apply)	`docx_prepare_context_model_name` / `code_generator_model_name`	🟡 Conditional — `requires_auth=false`; enabled per-agent; no hard env requirement
Tool Search	KMS models — Embedding Model	CapabilitySearchTool	`foundry_tool_embedding_model_name`	`BAAI/bge-large-en-v1.5` (embedding)	`foundry_tool_embedding_model_name`	🟡 Conditional — `always_available=false`; enabled per-agent
Meta Agent	Meta Agent (N — disabled by default)	PromptDraftingTool, ProposePatchTool	`app_config.tooling.meta_agent.llm_model_name` (toml)	`anthropic/claude-sonnet-4-6`	`app_config.tooling.meta_agent.llm_model_name`	🟡 Conditional — disabled by default (`meta_agent=@null` in `default.toml`); must be explicitly set in env toml to activate
Mufti Agent	—	KnowledgeSearchTool, ContentFilterTool, ResponseValidationTool	— (internal agent)	none	—	🟡 Conditional — `always_available=false`; internal sub-agent; enabled per-agent config
Obfuscation	—	PIISanitizerTool	— (rule-based)	none	—	🟡 Conditional — no dedicated MCP server; activated via agent/pipeline config
Time & Date	—	GetCurrentTimeTool	— (system clock)	none	—	🟡 Conditional — enabled per-agent; no credentials required
Persistent Object Mgmt	—	ObjectRetrievalTool	— (storage retrieval)	none	—	✅ Always on — `always_available=true` (foundational_tools server); no credentials required; available to all agents by default

Detailed Reference¶

Web Search¶

Server file: src/tooling/core/mcp_servers/servers/web_search.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
WebSearchTool, WebCrawlTool, ArxivSearchTool, ArxivUrlTool	`web_search_llm`	`azure/gpt-5-chat`	`llm_model_name`	🟡 Conditional — requires env var `WEB_SEARCH_TAVILY_API_KEY` (`strict=false`, degrades gracefully without it)

Code Execution¶

Server file: src/tooling/core/mcp_servers/servers/code_execution.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
CodeInterpreterExecutionTool, XlsxSkillTool, Daytona runtime/artifacts/helpers	— (no LLM call)	none	—	🟡 Conditional — `always_available=false`; enabled per-agent; no hard env requirement

Daytona Minimal Code Execution¶

Server file: bundled inside code_execution server

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
DaytonaMinimalCodeExecutionTool	— (no LLM call)	none	—	🟡 Conditional — bundled with code_execution server; same activation rule

Code Generator¶

Server file: src/tooling/core/mcp_servers/servers/code_generation_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
CodeGeneratorTool	`code_generator_llm`	`azure/gpt-5-chat`	`code_generator_model_name`	🟡 Conditional — `always_available=false`; also required by data_analysis as a `SERVER_DEPENDENCY`

Data Analysis¶

Server file: src/tooling/core/mcp_servers/servers/data_analysis.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
DataAnalysisTool (column_mapper, data_profiler, query_validator sub-utils)	`app_config.tooling.data_analysis.*` (dynaconf toml)	`azure/gpt-5-chat` (all 3 sub-models)	`app_config.tooling.data_analysis.{profiler,code_generator,column_mapper}_model`	🟡 Conditional — `always_available=false`; `SERVER_DEPENDENCY` on both `data_visualization` and `code_generation` servers being active

Production note: Sub-model config (profiler_model, code_generator_model, column_mapper_model) is set in config/default.toml under [default.tooling.data_analysis] — overridable via TOOLING__DATA_ANALYSIS__<FIELD> env vars (dynaconf double-underscore form).

Data Visualization¶

Server file: src/tooling/core/mcp_servers/servers/data_visualization.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
DataVisualizationTool	`data_visualization_llm`	`anthropic/claude-3-7-sonnet`	`data_visualization_llm`	🟡 Conditional — `always_available=false`; also a `SERVER_DEPENDENCY` for data_analysis

Production note: This is the only visualization tool configured with a Claude model by default (not GPT). Override via DATA_VISUALIZATION_LLM env var.

Image Analysis¶

Server file: src/tooling/core/mcp_servers/servers/vision_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
ImageAnalysisTool	`app_config.image_analysis.model` (toml)	`Qwen/Qwen3-VL-8B-Instruct` (VLM)	`app_config.image_analysis` (VLM config block)	🟡 Conditional — disabled by default (`image_analysis=@null` in `config/default.toml`); must be configured in environment-specific toml; uses `RequirementType.APP_CONFIG` gate

Production note: To enable, add to your env toml:

[environment.image_analysis]
model = "Qwen/Qwen3-VL-8B-Instruct"
timeout_seconds = 20
max_tokens = 1024
max_image_bytes = 5242880

Database Navigator¶

Server file: src/tooling/core/mcp_servers/servers/database_navigator.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
DatabaseNavigatorTool	`llm_model_name` / `small_llm_model_name` (fallback chain)	`azure/gpt-5-chat`	`llm_model_name` → `small_llm_model_name`	🟡 Conditional — `always_available=false`; no hard env requirement; enabled per-agent

PP DB Connector¶

Server file: src/tooling/core/mcp_servers/servers/pp_db_connector.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
ListDatabaseConnectionsTool, ExecuteDatabaseQueryTool, GetDatabaseConnectionTool	— (no LLM call)	none	—	🟡 Conditional — `requires_auth=false`; toggled by org/agent configuration

Knowledge Base¶

Server file: activated when KMS is attached to an agent

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
RetrievalTool (KMS chunks, documents, retrieve)	— (vector retrieval)	`BAAI/bge-large-en-v1.5` (embedding)	`foundry_tool_embedding_model_name`	🟡 Conditional — requires `KMS_BASE_URL`; enabled when a knowledge base is attached to an agent

Knowledge Extraction¶

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
DocumentKnowledgeExtractorTool	`extraction_llm`	`azure/gpt-5-chat`	`llm_model_name`	🟡 Conditional — requires `KMS_BASE_URL`; bundled with knowledge_base activation

Summarization¶

Server file: src/tooling/core/mcp_servers/servers/summarization_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
SummarizationTool	`summarization_llm`	`azure/gpt-5-chat`	`small_llm_model_name`	🟡 Conditional — `always_available=false`; enabled per-agent

Translation¶

Server file: src/tooling/core/mcp_servers/servers/translation.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
TextToTextTranslationTool, DocumentToDocumentTranslationTool	`translation_llm`	`azure/gpt-5-chat`	`llm_model_name`	🟡 Conditional — requires env var `KMS_BASE_URL` (`RequirementType.ENV_VAR`)

Email (Gmail)¶

Server file: src/tooling/core/mcp_servers/servers/email_workflow_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
ThreadSearchTool, FetchEmailsTool, FetchRelevantEmailsContentTool, SelectRelevantEmailsTool, ComposeAndExtractEmailTool, CreateUpdateDraftTool, SendGmailEmailTool, FetchAttachmentContentTool, EmailConsistencyCheckTool, BreakdownThreadFetchingQueryTool	`email_query_extraction_llm`, `thread_fetching_breakdown_llm`, `relevant_content_extraction_llm`, `extraction_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; user must complete Google OAuth flow (gmail scope); requires `GOOGLE_GMAIL_CLIENT_ID` + `GOOGLE_GMAIL_CLIENT_SECRET`

Outlook Email¶

Server file: src/tooling/core/mcp_servers/servers/outlook_email_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
SearchOutlookEmailsTool, FetchOutlookEmailsTool, SendOutlookEmailTool, CreateOutlookDraftTool	`email_query_extraction_llm`, `thread_fetching_breakdown_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; requires `MICROSOFT_CLIENT_ID` + `MICROSOFT_CLIENT_SECRET` + `MICROSOFT_TENANT_ID`

Outlook Calendar¶

Server file: src/tooling/core/mcp_servers/servers/outlook_calendar_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
FetchOutlookEventsTool, CreateOutlookEventTool, UpdateOutlookCalendarEventTool, CancelOutlookCalendarEventTool	`extraction_llm`, `people_query_breakdown_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; same Microsoft OAuth credentials as Outlook Email

Calendar (Google)¶

Server file: src/tooling/core/mcp_servers/servers/calendar_workflow_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
CreateCalendarEventTool, UpdateCalendarEventTool, CancelCalendarEventTool, ListCalendarSeriesTool, GetAvailabilityTool, CalendarConsistencyCheckTool, BreakdownCalendarQueryTool, ListCalendarsTool, ComposeAndExtractCalendarEventTool	`extraction_llm`, `thread_fetching_breakdown_llm`, `people_query_breakdown_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; requires `GOOGLE_CALENDAR_CLIENT_ID` + `GOOGLE_CALENDAR_CLIENT_SECRET`

Google Docs¶

Server file: src/tooling/core/mcp_servers/servers/docs_workflow_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
CreateGoogleDocTool, ReadGoogleDocTool, UpdateGoogleDocTool, DeleteGoogleDocTool, CopyGoogleDocTool, ListGoogleDocsTool	— (Google API calls)	none	—	🔐 OAuth required — `requires_auth=true`; requires `GOOGLE_DOCS_CLIENT_ID` + `GOOGLE_DOCS_CLIENT_SECRET`

Google Maps¶

Server file: src/tooling/core/mcp_servers/servers/google_maps.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
GoogleGeocodingTool, GoogleReverseGeocodingTool, GoogleDirectionsTool, GooglePlacesCategorySearchTool, MapVisualizationTool	— (Google Maps API)	none	—	🟡 Conditional — `always_available=false`; requires env var `GOOGLE_MAPS_API_KEY` (`RequirementType.ENV_VAR`)

OSM (OpenStreetMap)¶

Server file: src/tooling/core/mcp_servers/servers/open_street_maps.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
OSMGeocodingTool, OSMReverseGeocodingTool, OSMDirectionsTool, OSMCategorySearchTool, MapVisualizationTool	— (Nominatim API)	none	—	🟡 Conditional — requires env var `OSM_USER_AGENT` per Nominatim usage policy; no auth needed

People¶

Server file: bundled with email/calendar/docs workflow servers

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
SearchPeopleTool, EnrichPeopleTool, BreakdownPeopleQueryTool, LLM extract tools, Google Contacts enricher, Directory searcher	`people_extraction_llm`, `people_search_extraction_llm`, `people_query_breakdown_llm`	`azure/gpt-5-chat`	`llm_model_name`	🔐 OAuth required — `requires_auth=true`; activated together with email/calendar/docs servers

Online Media¶

Server file: src/tooling/core/mcp_servers/servers/youtube_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
YouTubeTool, YouTubePlaylistTool	— (YouTube API)	none	—	🟡 Conditional — requires env var `OXYLABS_PROXY_ENDPOINT` (`RequirementType.ENV_VAR`) for proxy-based YouTube access

DOCX¶

Server file: src/tooling/core/mcp_servers/servers/docx_workflow_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
DocxPrepareContextTool, DocxApplyOperationTool, DocxMarkdownToWordTool	`docx_prepare_context_llm`, `code_generator_llm`	`qwen/qwen3-5-122b-a10b` (prepare) / `azure/gpt-5-chat` (apply)	`docx_prepare_context_model_name` / `code_generator_model_name`	🟡 Conditional — `requires_auth=false`; enabled per-agent; no hard env requirement

Production note: DOCX is the only tool that uses two different models for two phases — the 120B model for context preparation (heavier reasoning) and the standard model for applying operations.

Tool Search¶

Server file: src/tooling/core/mcp_servers/servers/tool_search.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
CapabilitySearchTool	`foundry_tool_embedding_model_name`	`BAAI/bge-large-en-v1.5` (embedding)	`foundry_tool_embedding_model_name`	🟡 Conditional — `always_available=false`; enabled per-agent

Meta Agent¶

Server file: src/tooling/core/mcp_servers/servers/meta_agent.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
PromptDraftingTool, ProposePatchTool	`app_config.tooling.meta_agent.llm_model_name` (dynaconf toml)	`anthropic/claude-sonnet-4-6`	`app_config.tooling.meta_agent.llm_model_name`	🟡 Conditional — disabled by default (`meta_agent=@null` in `config/default.toml`); must be explicitly set in environment toml to activate

Production note: To enable, add to your env toml:
[environment.tooling.meta_agent]
llm_model_name = "anthropic/claude-sonnet-4-6"

Mufti Agent¶

Server file: src/tooling/core/mcp_servers/servers/mufti_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
KnowledgeSearchTool, ContentFilterTool, ResponseValidationTool	— (internal agent)	none	—	🟡 Conditional — `always_available=false`; internal sub-agent; enabled per-agent config

Persistent Object Management¶

Server file: src/tooling/core/mcp_servers/servers/foundational_tools.py

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
ObjectRetrievalTool	— (storage retrieval)	none	—	✅ Always on — `always_available=true` (foundational_tools server); no credentials required; available to all agents by default

Production note: This is the only tool that is always on. It is the sole member of the foundational_tools server. All other tools require some form of activation.

Obfuscation¶

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
PIISanitizerTool	— (rule-based)	none	—	🟡 Conditional — no dedicated MCP server; activated via agent/pipeline config

Time & Date¶

Individual Tools	Settings Property	Default Model	Central Settings Object	Enable / Disable
GetCurrentTimeTool	— (system clock)	none	—	🟡 Conditional — enabled per-agent; no credentials required

Summary¶

Stat	Count
Total tool categories	29
Categories with LLM calls	16
Categories with no LLM (pure API / rule-based)	13
Always on (no config needed)	1 (Persistent Object Mgmt)
OAuth required	5 (Gmail, Outlook Email, Outlook Calendar, Google Calendar, Google Docs)
Disabled by default in `default.toml`	2 (Image Analysis, Meta Agent)
Categories using Claude models	2 (Data Visualization, Meta Agent)
Categories using embedding models	2 (Knowledge Base, Tool Search)