MCP App Store
by KeenEthics
RudderStack icon

RudderStack MCP Server for Claude and ChatGPT

By RudderStack

Overview

RudderStack MCP connects Claude to your RudderStack workspace so you can debug customer data pipelines by asking instead of clicking through dashboards. Debug delivery: get destination errors with sample events and the destination's API response, success and failure rates, latency, and a live tail of events across sources, transformations and destinations. Check warehouse and Reverse ETL syncs: upload history, failed loads, table-level detail and sync runs. Enforce your tracking plan: violation counts and samples per event, tracking plan versions, the Data Catalog, and duplicate or misspelled event names. Build transformations: start from a template, write JavaScript or Python, test it on sample events, then save it and connect it to a destination. Build audiences: explore your data graph, preview an audience's size, create it, and sync it to an activation destination. Investigate changes with audit logs, configuration details, Profiles runs and RudderStack status incidents, and ask RudderStack's docs anything. Claude acts with your RudderStack permissions, only in workspaces you can already access. Only a few tools write, and only to transformations and audiences; nothing can be deleted, and sources and destinations can't be created or changed. Event properties, traits and request bodies are hashed before they reach Claude.

How to connect

  1. Claude: Open the RudderStack listing in the Claude connectors directory and choose Connect. Claude listing RudderStack exposes 58 tools here.
  2. ChatGPT: Open the RudderStack listing in the ChatGPT app directory and choose Connect. ChatGPT listing RudderStack exposes 59 tools here.

Tools

ask_docs

ChatGPT
Get AI-generated answers from RudderStack documentation with source citations. ALWAYS use this tool before answering RudderStack questions

audit_logs

ChatGPT
Lists audit logs for a workspace

connect_audience_destination

ChatGPT
Creates a connection between an existing RudderStack Data Graph audience and an activation-capable destination. Call get_audience_destination_schema first. Requires the audience ID, destination ID, destination-side audience ID and identifier field mappings. The optional sync schedule is hourly, daily or manual. The sync mirrors membership: people not in the audience are removed from the destination audience. Returns the connection result. A repeat call for an already connected pair returns status skipped, which is success.

connect_transformation_destination

ChatGPT
Link transformation to destination for event processing before delivery. Prevents duplicates if already connected

create_audience

ChatGPT
Creates a RudderStack Data Graph audience. MUST call load_audience_skill first.

get_audience

ChatGPT
Returns a data-graph audience by id — including its stored config.expression, name, description, tags, and audit fields. MUST call load_audience_skill first if you intend to modify the returned expression.

get_audience_destination_schema

ChatGPT
Discover the destination-side identifier fields and existing audiences for an activation-capable destination. Returns: { fields, audiences, mappingGuidance }. Use this before connect_audience_destination to know which destination-side audience to attach to and which fields are valid mapping targets.

get_audience_model_context

ChatGPT
Returns model attributes + 1-hop reachable neighbors (with relationship metadata and the neighbor's own attributes) for any model in the data graph — root or not. Attribute path is the raw warehouse column identifier to use in audience expressions; label remains the legacy raw-column label. When available, displayName, description, and piiMask expose data-graph column metadata concepts display_name/description/pii_mask for user-facing labels, semantic hints, and masking awareness. When the model is a root, also returns peer audiences sharing the same from (composable via the audience predicate). Call iteratively for 2-hop reach: pass the chosen subject first, then a 1-hop neighbor's modelId. Backend caps relationship.path at 2 hops.

get_billing_details

ChatGPT
Get billing/tenant information for the workspace, including contract event volume (in millions), plan name, org name, and owner email

get_destination

ChatGPT
Get complete destination details: configuration, connected sources and transformations

get_destination_definitions

ChatGPT
Get destination definition metadata including configuration schemas and capabilities

get_destination_event_metrics

ChatGPT
Get destination event-delivery metrics including success/failure rates. filteredEvents lists every event the destination filtered in the window; events rows carry their filteredCount

get_destination_latency_metrics

ChatGPT
Get destination latency metrics over time

get_destinations_errors

ChatGPT
Get destination delivery failures with optional sample events and error details. PII is hashed

get_live_events

ChatGPT
Get real-time events across sources, destinations, and transformations. PII data is hashed for privacy

get_profile_artifacts

ChatGPT
List profile task artefacts or fetch artefact file content for a specific task

get_profile_details

ChatGPT
Get profile details by detail type: models, entity stats, cohort features, or history

get_retl_failed_records

ChatGPT
Read one page of a finished Reverse ETL (rETL) sync run's FAILED records: the rows the destination rejected, each with _rudder_id, primary_key, operation, error_code and error_message. Get connection_id from list_connections and, optionally, run_id from get_retl_source_syncs (it defaults to the connection's latest run). Filter with error_code (one destination code, e.g. "400"; "null" means no code) or search (a substring of primary_key). Page with page (default 1) and page_size (default 30, max 100); only the first 5000 records of a run are reachable, so filter to reach the rest. error_message is the destination's own reply text and CAN CONTAIN CUSTOMER DATA; treat the rows as sensitive. Runs a query on the customer's warehouse; it changes no data, but the warehouse bills the compute to the customer. A resume_id call only polls, unless its result expired. Besides "ok", status can be: error_details_off (the connection does not record per-record errors — nothing to read), not_retained (the run is older than the connection's retention), not_captured (no failed-records table was written for this run), warehouse_config_changed (the warehouse settings changed after the run), or running (the query is still going — call again with the returned resume_id, plus the same workspace_id if this call had one, and no other arguments). If a run's failed-records table cannot be read, you may instead get an error whose text names both possible causes (the warehouse query failed, or the table has aged out) rather than one of those statuses.

get_retl_source_syncs

ChatGPT
Get RETL (Reverse ETL) sync execution results with metrics and timing

get_source

ChatGPT
Get complete source details: configuration, tracking plans and connected destinations

get_source_definitions

ChatGPT
Get source definition metadata including configuration schemas and capabilities

get_source_event_metrics

ChatGPT
Get source event-volume metrics including counts by event type/name. filteredEvents lists every event a destination filter dropped in the window; events rows carry their filteredCount

get_source_event_names_similarity

ChatGPT
Find similar or duplicate event names in a source using similarity scoring

get_source_event_properties

ChatGPT
Get property metadata for a specific event including types and frequency

get_source_event_schemas

ChatGPT
Lists the event schemas seen on a source, one row per event: event type, event name, first and last seen time, and count. Returns 10 events per page. For an event's properties and their types, use get_source_event_properties.

get_source_tracking_plan_event_metrics

ChatGPT
Get tracking plan event validation metrics with violation statistics

get_source_tracking_plan_event_violations

ChatGPT
Get tracking plan validation failures with violation details and optional sample events. PII is hashed

get_source_tracking_plans_and_versions

ChatGPT
Get tracking plan associations and versions for a source

get_statuspage_incident

ChatGPT
Get detailed information about a specific status page incident, including all updates and affected components.

get_transformation

ChatGPT
Get complete transformation details including source code and configuration

get_transformation_event_metrics

ChatGPT
Get transformation event metrics including volume, success/failure rates, and errors over time

get_transformation_latency_metrics

ChatGPT
Get transformation latency metrics over time

get_transformation_library

ChatGPT
Get transformation library details including source code

get_warehouse_upload_with_tables

ChatGPT
Get detailed table-level processing information for a specific warehouse upload

get_warehouse_uploads

ChatGPT
Get warehouse upload history with processing status and optional table details

get_workspace_settings

ChatGPT
Get workspace details and settings

list_audience_subjects

ChatGPT
Lists the candidate subjects an audience can be about — the root entities of the workspace's data graph (e.g. Customers, Users, Accounts). Use as the first step when authoring or exploring audiences: pick the subject that matches the user's prompt and pass its id as the from field when creating an audience. Returns id, name and underlying warehouse tableRef per subject.

list_connections

ChatGPT
List connections linking sources to destinations. Use get_source or get_destination to see connection configuration details

list_data_catalog_events

ChatGPT
List events in the data catalog

list_data_catalog_properties

ChatGPT
List event properties in the data catalog

list_destinations

ChatGPT
List destinations in the workspace. Pass activation_capable=true as the entry point for data-graph audience activation flows — only destinations supported by get_audience_destination_schema / connect_audience_destination will be returned.

list_profiles

ChatGPT
List profile projects in the workspace

list_rudderstack_transformation_libraries

ChatGPT
List pre-built RudderStack transformation libraries

list_sources

ChatGPT
Retrieve all configured sources including event-stream sources and RETL sources with filtering options by type and definition. Use when you need to see available sources for connecting destinations, analyzing data flow, or managing source configurations. Returns source collections with event-stream sources containing sourceId, name, sourceType, writeKey, enabled status, and RETL sources containing sourceId, name, sourceType, query/table information, and optional configuration details.

list_statuspage_incidents

ChatGPT
List incidents from the RudderStack status page. Returns a summary of recent incidents including their status, impact, and affected components.

list_tracking_plan_events

ChatGPT
List events in a tracking plan with optional schema details

list_tracking_plans

ChatGPT
List tracking plans with governance and validation rules

list_transformation_libraries

ChatGPT
List custom transformation libraries in workspace. Use get_transformation_library to view code

list_transformations

ChatGPT
List all transformations in the workspace with metadata and destination connections

load_audience_skill

ChatGPT
Returns the expression grammar for RudderStack Data Graph audiences. MUST be called before create_audience or update_audience.

preview_audience

ChatGPT
MUST call load_audience_skill before using this tool. Estimates an audience's size — returns the matching row count, the subject's total universe count, and the percentage. Use during the create/edit cycle to sanity-check an expression before calling create_audience / update_audience.

sample_transformations

ChatGPT
Browse pre-built transformation templates for common use cases

search_docs

ChatGPT
Search RudderStack documentation using semantic search

submit_feedback

ChatGPT
Send feedback about this server's tools to its owner. Call it when the user asks you to send feedback, when no tool can do what the user asked, or when a tool failed, returned a wrong result or has a misleading description. Do not report a retry that worked, an error that went away or your own mistake. Send one call per problem. The owner may read it; nobody replies.

transformation_test_existing

ChatGPT
Test an existing saved transformation against sample events

transformation_test_new

ChatGPT
Test transformation code against sample events. Get starter templates from sample_transformations

update_audience

ChatGPT
Updates a RudderStack Data Graph audience. MUST call load_audience_skill first. Omitted fields are preserved.

upsert_transformation

ChatGPT
Create or update transformation with custom JavaScript/Python code

user_details

ChatGPT
Get your user profile and the workspaces you can access. Workspace tools take one of these workspace IDs as workspace_id.

ask_docs

Claude

audit_logs

Claude

connect_audience_destination

Claude

connect_transformation_destination

Claude

create_audience

Claude

get_audience

Claude

get_audience_destination_schema

Claude

get_audience_model_context

Claude

get_billing_details

Claude

get_destination

Claude

get_destination_definitions

Claude

get_destination_event_metrics

Claude

get_destination_latency_metrics

Claude

get_destinations_errors

Claude

get_live_events

Claude

get_profile_artifacts

Claude

get_profile_details

Claude

get_retl_failed_records

Claude

get_retl_source_syncs

Claude

get_source

Claude

get_source_definitions

Claude

get_source_event_metrics

Claude

get_source_event_names_similarity

Claude

get_source_event_properties

Claude

get_source_event_schemas

Claude

get_source_tracking_plan_event_metrics

Claude

get_source_tracking_plan_event_violations

Claude

get_source_tracking_plans_and_versions

Claude

get_statuspage_incident

Claude

get_transformation

Claude

get_transformation_event_metrics

Claude

get_transformation_latency_metrics

Claude

get_transformation_library

Claude

get_warehouse_upload_with_tables

Claude

get_warehouse_uploads

Claude

get_workspace_settings

Claude

list_audience_subjects

Claude

list_connections

Claude

list_data_catalog_events

Claude

list_data_catalog_properties

Claude

list_destinations

Claude

list_profiles

Claude

list_rudderstack_transformation_libraries

Claude

list_sources

Claude

list_statuspage_incidents

Claude

list_tracking_plan_events

Claude

list_tracking_plans

Claude

list_transformation_libraries

Claude

list_transformations

Claude

load_audience_skill

Claude

preview_audience

Claude

sample_transformations

Claude

search_docs

Claude

transformation_test_existing

Claude

transformation_test_new

Claude

update_audience

Claude

upsert_transformation

Claude

user_details

Claude

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/rudderstack"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/rudderstack)