Skip to main content

MCP Tools Reference

All 34 tools available via the MCP endpoint. Each tool accepts JSON parameters and returns structured responses.

Curated Tools (24)

Execute a Google Ads Query Language (GAQL) query.

hubspot_crm

Full CRUD on HubSpot CRM objects.

salesforce_query

Execute SOQL queries against Salesforce.

salesforce_crm

Insert, update, or describe Salesforce objects.

ga4_report

Run Google Analytics 4 reports.

gsc_performance

Query Google Search Console performance data.

semrush_research

SEMrush domain analytics and keyword research.

gmail

Search, read, or send Gmail messages.

google_calendar

Google Calendar event CRUD. Web-grounded search via Perplexity Sonar models.

aws_bedrock_invoke

Invoke AWS Bedrock models (raw API).

aws_bedrock_converse

Bedrock Converse API with tool use support.

aws_textract_analyze

Extract text and tables from PDFs and images.

aws_ses_send

Send transactional emails via AWS SES.

aws_nova_canvas

Generate images via AWS Nova Canvas.

gamma_generate

Generate presentations and documents via Gamma.

linkedin_ads

LinkedIn campaign and ad management.

meta_ads

Meta (Facebook/Instagram) Ads API.

microsoft_ads

Microsoft Advertising (Bing Ads) API.

microsoft_calendar

Microsoft 365 Calendar operations.

dealcloud

DealCloud CRM operations (PE/VC).

submit_feedback

Submit quality feedback on a tool call output.

web_fetch

Fetch and parse a web page.

search_knowledge

Search the tenant knowledge base (Vectorize index).

Platform Tools (10)

agent_state

Store and retrieve key-value state scoped to the current tenant.

batch_execute

Execute up to 10 API calls in parallel.

discover_apis

Returns all connected providers with example endpoints. No parameters required.

call_api

Generic authenticated proxy for any registered domain.

context_query

Query the context plane for facts, signals, and entity data.

context_explain

Explain context plane reasoning and signal evaluations.

ai_invoke

Route LLM calls through the AI Gateway with automatic model selection, observability, and budget cap. Provider and model are chosen automatically based on intent and quality — do not specify them directly.

llm_invoke

Direct LLM invocation with explicit provider and model selection, or shortcut tier routing.

gemini_invoke

Invoke Google Gemini models directly.

claude

Invoke Claude models for nested agent calls, batch processing, file operations, or sub-agent spawning. Uses a discriminated union on action. action: "invoke" (most common) Other action values: "batch", "files", "agents" — each has its own parameter set.

Common Parameters

All curated tools support: