The Maven MCP server

The core insight surface. Assistants discover metrics and tables, then query the governed semantic layer — never raw SQL.

• query_metrics

  Query workspace analytics across every connected domain — ads, web analytics, attribution, CRM, ecommerce, email, retail POS, SEO, and social.

  Try: “Show me last 30 days of revenue by platform.”

• get_dimension_values

  List the real distinct values of a dimension in this workspace — platform ids, stage labels, campaign names — before filtering.

  Try: “What campaign names are running in Google Ads this month?”

• list_metrics

  List every metric in the semantic catalog across all domains.

  Try: “What metrics can I query for email marketing?”

• describe_schema

  Discover queryable tables — canonical domain, grain, time field, and connected sources — or inspect one table’s full semantic contract.

  Try: “What tables do I have, and what does each one cover?”

• compare_periods

  Compare a metric across two time periods and return the delta and percent change.

  Try: “Compare ad spend this month vs last month.”

• top_n

  Return the top (or bottom) N values of a metric grouped by a dimension over a time range.

  Try: “Top 10 campaigns by ROAS last quarter.”

Governed event trends, funnels, retention, paths, lifecycle, and stickiness with pseudonymous actor drill-downs.

• describe_behavior_schema

  Describe provider capabilities, discovered events, governed properties, and approved group types before querying.

  Try: “Which product events and properties are ready to analyze?”

• query_behavior_trends

  Query event trends with actor, session, or group counts, governed breakdowns, and numeric property aggregates.

  Try: “Trend weekly activated users by plan for the last quarter.”

• query_behavior_funnel

  Analyze 2–8 step funnels with ordered, strict, or unordered sequence semantics and bounded conversion windows.

  Try: “Show signup-to-activation conversion within seven days.”

• query_behavior_funnel_actors

  Return the pseudonymous actors that completed a governed funnel without exposing provider IDs or person properties.

  Try: “Which anonymous actors completed the activation funnel?”

• query_behavior_retention

  Build first-time or recurring retention cohorts over daily, weekly, or monthly periods.

  Try: “What is eight-week retention after first activation?”

• query_behavior_retention_actors

  Return pseudonymous retained actors for a governed retention query.

  Try: “Which anonymous actors returned in week four?”

• query_behavior_paths

  Find common event paths with repeated-node collapsing, bounded depth, and minimum-count pruning.

  Try: “What paths do users take before upgrading?”

• query_behavior_path_actors

  Return pseudonymous actor keys grouped into common behavioral paths.

  Try: “Which anonymous actors followed the checkout-abandonment path?”

• query_behavior_lifecycle

  Classify new, returning, resurrecting, and dormant actors by activity interval.

  Try: “How many users became dormant this month?”

• query_behavior_stickiness

  Measure exact or cumulative active-interval distributions with an optional event threshold.

  Try: “How many users were active on at least four days this month?”

• get_behavior_query_result

  Retrieve status or results for a longer-running behavioral query job.

  Try: “Get the result of the retention query that is still processing.”

• list_behavior_group_mappings

  List suggested, active, or dismissed provider-property mappings to canonical behavior groups.

  Try: “Which account and organization mappings have been discovered?”

• promote_behavior_group_mapping

  Admin-only: approve a provider property as a canonical group type.

  Try: “Approve organization_id as the canonical account group.”

• dismiss_behavior_group_mapping

  Admin-only: dismiss a suggested or active behavior group mapping.

  Try: “Dismiss the noisy workspace_label mapping.”

Read-only account journeys, deal influence, intent signals, and normalized campaign control-plane resources.

• list_account_signals

  List recent warm-account engagement and intent signals enriched with CRM account and owner identity.

  Try: “Which accounts are showing buying intent this week?”

• get_account_journey

  Return the chronological marketing-engagement and CRM outcome journey for one canonical account.

  Try: “Show the full journey for Acme before its opportunity opened.”

• explain_deal_influence

  Explain whether one deal was influenced and show deduplicated account-level campaign evidence.

  Try: “Which campaigns influenced this deal, and why?”

• inspect_campaign_resources

  Inspect normalized campaign resources from Maven’s persisted provider inventory without a live provider call.

  Try: “List active and paused campaigns with their URL settings.”

Curated, warehouse-backed store records and read-only retail decisions. PII is masked unless the caller is an admin.

• describe_commerce_records

  Describe the curated commerce entities, allowed fields, filters, date limits, pagination, and PII behavior.

  Try: “What commerce entities and fields can I query?”

• query_commerce_records

  Page through transactions, line items, products, customers, inventory, fulfillments, returns, payments, promotions, and price rules.

  Try: “Show me yesterday’s transactions with their line items.”

• get_commerce_record

  Fetch one curated commerce record by entity and source record ID.

  Try: “Get the full record for Shopify order #4317.”

• forecast_commerce_demand

  Read-only 14-day unit forecast from trailing 28-day sales, with days of cover.

  Try: “Which products will run out of stock in the next two weeks?”

• recommend_commerce_restock

  Read-only restock recommendations from current inventory, incoming units, reorder levels, and forecasted demand.

  Try: “What should we reorder this week?”

• recommend_commerce_transfers

  Read-only KORONA location-to-location rebalancing candidates — never creates a transfer.

  Try: “Which stores should transfer stock to cover the downtown location?”

• preview_commerce_identity_override

  Admin-only dry run for linking one Shopify product or customer to one KORONA record, with the current match state.

  Try: “Check what linking this Shopify customer to KORONA record 812 would do.”

• apply_commerce_identity_override

  Admin-only audited write that applies a previously reviewed Shopify↔KORONA identity link.

  Try: “Apply the identity link we just previewed.”

• remove_commerce_identity_override

  Admin-only soft removal of a commerce identity override; audit history is retained.

  Try: “Remove the override on SKU 1042.”

Manage the workspace itself: connected platforms, sync runs, freshness, team members, and budgets.

• whoami

  Show the current credential type, scopes, and connector-local authorization context — without echoing the raw key.

  Try: “What can this connection access?”

• list_capabilities

  List the capability packs available to this connection, including which are open and which need another scope.

  Try: “What else can this Maven connection do?”

• open_capability

  Open one capability pack for the current session and return its available tool names and summaries.

  Try: “Open the tracking and attribution capability.”

• list_workspaces

  List every workspace accessible through this connection — one for workspace keys, all for agency credentials.

  Try: “Which of my client workspaces can I query?”

• get_workspace_info

  Workspace metadata: name, plan, dataset, active integration count, and last model run.

  Try: “What plan is this workspace on, and when did the models last build?”

• get_workspace_brief

  The first-call intelligence brief: context, connected platforms, coverage, readiness, caveats, and next actions.

  Try: “Give me the lay of the land for this workspace.”

• list_platforms

  List every data platform a Maven workspace can connect.

  Try: “Which CRMs does Maven support?”

• recommend_platforms

  Recommend (or record) the first data sources to connect, based on business model, industry, and goals.

  Try: “We’re a DTC brand on Shopify spending on Meta and Google — what should we connect first?”

• create_workspace

  Create a new workspace under the current customer account (agency-scoped credentials only).

  Try: “Spin up a workspace for our new client, Acme Co.”

• list_integrations

  List connected integrations with capability status, freshness, and remediation.

  Try: “Which ad platforms are connected right now?”

• get_integration

  Fetch details for one specific integration by UUID.

  Try: “What’s the status of our Google Ads connection?”

• connect_integration

  Generate a shareable connect link for OAuth platforms, or return credential requirements.

  Try: “Get me a link to connect TikTok Ads.”

• disconnect_integration

  Revoke a platform connection and stop its syncs.

  Try: “Disconnect the old Facebook Ads account.”

• list_runs

  List recent sync runs with status, timing, and errors.

  Try: “Why is HubSpot data stale — did the last sync fail?”

• trigger_sync

  Manually trigger an immediate data sync for one integration.

  Try: “Sync Google Ads now.”

• check_freshness

  Per-platform data freshness: status, last successful sync, and fallbacks.

  Try: “Is all of today’s data in yet?”

• list_users

  List workspace members with email, role, and join date.

  Try: “Who has access to this workspace?”

• invite_user

  Invite one person to the workspace by email — viewer access by default.

  Try: “Invite sarah@acme.com as a viewer.”

• add_workspace_viewers

  Bulk-grant viewer access across one or more workspaces.

  Try: “Give the three new analysts viewer access to all Acme workspaces.”

• get_media_budget_pacing

  Monthly media budget pacing: targets, spend-to-date, and projected month-end spend.

  Try: “Are we pacing to hit this month’s media budget?”

Backtested metric alerts in Slack or email, plus recurring PDF/PNG delivery of reports and decks.

• preview_alert

  Backtest an alert rule against recent data before anything is created.

  Try: “What would an alert for ROAS below 2.0 have caught last month?”

• preview_alert_in_channel

  Post a live preview of an alert to the workspace Slack channel — still no rule created.

  Try: “Show the team what that spend alert would look like in Slack.”

• create_alert

  Create and schedule an alert rule after the preview is approved.

  Try: “Alert me in Slack whenever daily spend exceeds $5,000.”

• update_alert

  Update, pause, resume, or edit an existing alert rule.

  Try: “Pause the CPA alert while we’re in the promo window.”

• delete_alert

  Delete an alert rule.

  Try: “Delete the old budget alert.”

• list_alerts

  List alert rules with last-evaluated status.

  Try: “Which alerts are active right now?”

• schedule_delivery

  Schedule recurring email delivery of a report or deck, sent as PDF and/or PNG.

  Try: “Email the exec summary to the board every Monday at 8am.”

• list_deliveries

  List scheduled deliveries with cadence, artifact, and recipients.

  Try: “What recurring report emails are set up?”

• update_delivery

  Edit a scheduled delivery’s recipients, cadence, formats, or enabled state.

  Try: “Add the new CMO to the weekly deck email.”

• delete_delivery

  Delete a scheduled delivery and stop all future sends.

  Try: “Stop the Friday PDF email.”

• list_pending_actions

  List staged agent writes awaiting a decision, including retryable partial-failure bundles.

  Try: “Which conversion or GTM changes are waiting for approval?”

• decide_pending_action

  Approve or reject one staged action; approval executes only after explicit user confirmation.

  Try: “Approve the reviewed GTM publish request.”

Find active authoring drafts and recent managed report work for the current connection.

• list_report_work

  Recovery surface: list active Report Studio drafts and recent managed builds for this connection.

  Try: “Where was I on that report draft?”

Persistent workspace memory plus certified metrics, saved views, and CRM custom fields — the trust layer assistants read first.

• get_workspace_context

  Read the full memory bundle: semantic facts, top-ranked notes, and certified metrics and views.

  Try: “What do we know about this business before I analyze?”

• update_workspace_context

  Update structured identity fields: business type, industry, primary KPIs, fiscal year start, timezone, and currency.

  Try: “Our fiscal year starts in July — update the workspace.”

• add_note

  Persist a free-form note about the workspace or customer.

  Try: “Remember that the May spike was a one-off PR mention.”

• list_notes

  Search workspace memory notes by category, date, keyword, or tags.

  Try: “What notes do we have about attribution?”

• supersede_note

  Replace a stale note; the old version stays archived for audit.

  Try: “Update the note about our primary KPI — it’s now contribution margin.”

• certify_metric

  Create a workspace metric definition that overrides the global catalog (admin; formulas reference catalog metrics only).

  Try: “Certify our internal ROAS definition that excludes brand campaigns.”

• list_custom_fields

  List the CRM custom fields discovered in this workspace’s synced data.

  Try: “Which HubSpot custom properties came through the sync?”

• promote_custom_field

  Make a discovered CRM custom field queryable as a dimension or metric (admin).

  Try: “Make the lifecycle_stage_v2 field queryable.”

• dismiss_custom_field

  Hide a junk or sensitive CRM custom field from the queryable set (admin).

  Try: “Hide the internal_notes field from queries.”

• certify_view

  Save a certified (admin) or personal saved query template for repeatable reporting.

  Try: “Save this as our weekly pacing view.”

• list_certified

  List all certified metrics and saved/certified views for the workspace.

  Try: “What are our approved standard reports?”

• use_certified_view

  Execute a saved or certified view by name.

  Try: “Run the executive summary view.”

Free-form context documents per customer, workspace, or data source — what the numbers mean, maintained over time.

• get_context

  Read the narrative context document (agent summary + user input) and recent changelog for a scope.

  Try: “Read the narrative context for the Google Ads source.”

• update_context

  Write the agent-maintained summary or user input for a context document.

  Try: “Summarize what we learned about Q2 creative performance.”

• append_context_changelog

  Append a dated changelog entry to a context document.

  Try: “Log that we switched attribution windows on June 1.”

• list_context_gaps

  Report where narrative context is missing across the customer.

  Try: “Which sources still have no context documented?”

• search_context

  Search governed notes, facts, decisions, source evidence, and active website excerpts with tenant authorization applied before ranking.

  Try: “What have we documented about the revenue definition and its history?”

• list_site_contexts

  List website context sources with build status, coverage, freshness, and stable site IDs.

  Try: “Which websites have a current governed context build?”

• get_site_context

  Read bounded website context with page provenance, scan time, and confidence — never raw HTML.

  Try: “Summarize the pricing and conversion surfaces on our primary site.”

• search_site_context

  Search sanitized website evidence for company, page, template, conversion, and implementation questions.

  Try: “Where does the site describe enterprise pricing?”

• refresh_site_context

  Queue an asynchronous website-context refresh while the current active context stays available.

  Try: “Refresh the primary site context after today’s launch.”

Freshness, anomaly, and tagging diagnostics — with remediation proposals that never write without human approval.

• data_quality_report

  Freshness and anomaly status from the workspace data-quality tables.

  Try: “Is the data clean enough to report on today?”

• get_utm_coverage

  UTM tagging coverage by platform — missing canonical IDs or misaligned tags over the last 30 days.

  Try: “How much of our paid traffic is missing UTMs?”

• ga4_not_set_report

  Diagnose GA4 “(not set)” traffic volume and how much of it is recoverable.

  Try: “Why is 20% of GA4 traffic “(not set)”?”

• get_attribution_inventory

  Read-only provider control-plane inventory, including paused and zero-spend entities with inherited URL settings.

  Try: “List every campaign in our ad accounts, even paused ones.”

• preview_utm_remediation

  Deterministic before/after UTM remediation proposal for one entity — never mutates the provider account.

  Try: “What would fixing UTMs on that campaign look like?”

• request_utm_writeback

  Queue a previewed UTM remediation proposal for human approval.

  Try: “Send that UTM fix for approval.”

Container audits and live-page QA, plus staged container changes that go live only after human approval.

• gtm_audit

  Static, read-only audit of the synced GTM container config — tags, triggers, variables, and folders.

  Try: “Audit our GTM container for missing conversion tags.”

• gtm_behavioral_qa

  Verify on a live page, in a real browser, that GTM loads and tags actually fire.

  Try: “Check that the purchase tag fires on the thank-you page.”

• gtm_consent_qa

  Diagnose baseline, accepted, and rejected consent states in a real browser and compare expected vendor behavior.

  Try: “Verify GA4 and Meta respect consent on this checkout page.”

• gtm_tagging_targets

  Discover stable live-page selectors for forms, contact links, CTAs, and embedded widgets before staging tags.

  Try: “Find reliable selectors for the demo form and phone links.”

• gtm_prepare_workspace

  Create a clean, health-gated GTM workspace for staging changes.

  Try: “Prepare a GTM workspace for the new pixel rollout.”

• gtm_stage_change

  Create, update, or delete tags, triggers, or variables in the staged (non-live) workspace.

  Try: “Stage a new GA4 event tag for signup.”

• gtm_stage_consent_repair

  Stage narrowly scoped additional-consent or Consent Initialization repairs in a non-live GTM workspace.

  Try: “Stage the consent-mode repair identified by the QA run.”

• gtm_request_publish

  Queue staged GTM changes for human approval — never publishes directly.

  Try: “Request approval to publish the staged tags.”

First-party attribution setup and governed conversion configuration — every provider write is queued for human approval.

• provision_site_key

  Get-or-create the workspace’s first-party attribution site key (idempotent).

  Try: “Set up first-party tracking on our site.”

• tracking_config

  Load cross-platform conversion configuration from synced tables — including GA4 key events and Create/Modify Event rules.

  Try: “What conversions do we have configured across GA4, Google Ads, and Meta?”

• ga4_access_check

  Preflight GA4 Admin API access before proposing any changes.

  Try: “Can Maven actually write to our GA4 property?”

• tracking_request_conversion_setup

  Preflight and queue one governed conversion setup across Google Ads, Meta, LinkedIn, TikTok, and StackAdapt.

  Try: “Set up the purchase conversion everywhere.”

• tracking_readiness_map

  Turn a tracking ticket and target pages into an evidence-backed map of required events, destinations, gaps, and next actions.

  Try: “Map this signup tracking request against what is live today.”

• ads_request_conversion_action

  Queue a Google Ads conversion action for human approval.

  Try: “Create a Google Ads conversion for demo bookings.”

• meta_request_conversion_action

  Queue a Meta custom conversion for human approval.

  Try: “Create a Meta custom conversion for checkout starts.”

• ga4_request_key_event

  Queue a GA4 key event (conversion) for human approval.

  Try: “Mark generate_lead as a key event in GA4.”

• ga4_request_custom_dimension

  Queue a GA4 custom dimension for human approval.

  Try: “Register mm_visitor_id as a GA4 custom dimension.”

• get_tracking_readiness

  Return approved tracking expectations and independent placement, configuration, runtime, delivery, and consent evidence.

  Try: “Is our approved purchase tracking ready end to end?”

Cross-workspace comparisons, anomaly sweeps, and reusable named analyses.

• benchmark_metric

  Compare one workspace’s metric against peer workspaces under the same customer account.

  Try: “How does Acme’s CTR compare to our other clients?”

• scan_anomalies

  Proactively scan a workspace — or all workspaces under a customer key — for unusual metric values.

  Try: “Scan all my clients for weird spend spikes this week.”

• save_analysis

  Save a query as a named, reusable analysis.

  Try: “Save this as “monthly channel mix”.”

• run_saved_analysis

  Execute a saved or certified analysis by name.

  Try: “Run the monthly channel mix analysis.”

Readiness, observational regional response, and audited matched-market geo-lift experiments at state or region grain.

• get_geo_measurement_readiness

  Check whether conformed regional outcomes, market coverage, and paid-media spend can support geo measurement.

  Try: “Are we ready to design a state-level geo-lift test?”

• analyze_geo_campaign_response

  Analyze observational regional response after a campaign launch, clearly separated from causal lift.

  Try: “How did regional outcomes move after the Texas campaign launched?”

• list_geo_experiments

  List causal geo experiments, optionally filtered by lifecycle status.

  Try: “Which geo experiments are currently running?”

• get_geo_experiment

  Get one experiment’s markets, launch playbook, compliance, counterfactual series, and latest lift readout.

  Try: “Show the current readout for the Midwest test.”

• design_geo_experiment

  Create an audited draft matched-market design after explicit confirmation.

  Try: “Design a six-week revenue geo test for Meta campaigns.”

• start_geo_experiment

  Start a reviewed geo experiment with an append-only actor audit event.

  Try: “Start the approved Midwest experiment.”

• analyze_geo_experiment

  Run the causal readout for an experiment with explicit confirmation and audit correlation.

  Try: “Analyze lift for the completed geo experiment.”

• cancel_geo_experiment

  Cancel one geo experiment after explicit confirmation.

  Try: “Cancel the experiment because the media plan changed.”

The canonical brand theme, readable by any assistant producing content.

• get_design_context

  Read the workspace’s central design file — brand palette, data-viz series colors, and font stack.

  Try: “What brand colors should this chart use?”

A direct line from your assistant to the Maven development team.

• submit_bug

  File a Maven platform bug with the development team — creates a tracked issue and returns its URL.

  Try: “Report a bug: the freshness check returns stale timestamps.”

• request_help

  Create a confirmed bug or support request with useful workspace and trace context.

  Try: “Open a support request for this failed connector with the current trace.”

The full draft-to-publish report authoring loop, mounted as an interactive MCP App. OAuth only — per-user identity is required.

• start_report_authoringOAuth only

  Start a durable report draft; returns the session_id every other draft tool needs.

  Try: “Build me a Q2 board report from our data.”

• open_report_for_editOAuth only

  Copy the current published report into an owner-scoped draft so edits publish as a new immutable version.

  Try: “Open the latest QBR report so I can update it.”

• import_report_artifactOAuth only

  Hand a finished, self-contained HTML or React artifact to Maven wholesale in one call.

  Try: “Import this HTML dashboard as a Maven report.”

• list_report_draftsOAuth only

  List the signed-in user’s active, non-expired report drafts.

  Try: “Which report drafts do I have open?”

• get_report_authoring_contextOAuth only

  Load the deterministic design.md, runtime contract, and semantic schema for authoring.

  Try: “What data can I bind into this report?”

• list_report_filesOAuth only

  List source files and byte sizes in a draft, with the current revision.

  Try: “What files are in this draft?”

• read_report_fileOAuth only

  Read one source file from a draft before editing.

  Try: “Show me the current index.html.”

• write_report_fileOAuth only

  Create or replace one report source file, guarded by optimistic revision checks.

  Try: “Rewrite the hero section with the new KPIs.”

• edit_report_fileOAuth only

  Apply an exact string replacement to one report source file.

  Try: “Change the title to “Q2 Board Review”.”

• delete_report_fileOAuth only

  Delete one source file using revision compare-and-swap.

  Try: “Remove the unused appendix file.”

• upsert_report_bindingOAuth only

  Create or replace a live Maven data binding without hand-authoring spec JSON.

  Try: “Bind the revenue chart to the gross_revenue metric.”

• delete_report_bindingOAuth only

  Delete a report data binding by key.

  Try: “Remove the stale spend binding.”

• probe_report_dataOAuth only

  Dry-run every declared binding through the real resolver — readiness, row counts, and freshness.

  Try: “Does every chart on this report have live data?”

• validate_reportOAuth only

  Compile and validate the current source revision without publishing; failures list numbered issues.

  Try: “Check this draft for errors.”

• preview_reportOAuth only

  Render the draft with real workspace data in the interactive Report Studio app — no publish.

  Try: “Show me how the report looks with real data.”

• publish_reportOAuth only

  Atomically publish the exact draft revision as a native Maven report.

  Try: “Publish it for the team.”

• abandon_report_draftOAuth only

  Soft-abandon an active draft; published report lineage is preserved.

  Try: “Scrap this draft.”

• open_reportOAuth only

  Return the authenticated native Maven URL for a published report.

  Try: “Open the published board report in Maven.”

• set_report_visibilityOAuth only

  Change an owned report between private and workspace visibility.

  Try: “Share the report with the whole workspace.”

• refresh_report_frameOAuth only

  Refresh the mounted Report Studio preview; does not publish.

  Try: “Reload the preview with the latest data.”

Customer-wide brand themes, published through a propose-then-confirm flow. OAuth only, admin scope.

• get_customer_brand_themeOAuth only

  Read the customer-wide published brand theme and any unpublished draft.

  Try: “What’s our published brand theme?”

• propose_customer_brand_theme_updateOAuth only

  Create an unpublished, immutable theme proposal by applying a narrow patch to the current theme.

  Try: “Propose updating the primary color to the new brand teal.”

• publish_customer_brand_theme_updateOAuth only

  Publish the exact current theme proposal after explicit user confirmation.

  Try: “Publish the theme update we just reviewed.”

• refresh_report_brand_contextOAuth only

  Refresh an owned Report Studio session to the latest published customer brand theme.

  Try: “Re-skin my draft with the latest brand theme.”

Import finished Claude/ChatGPT artifacts — slide decks and workbooks — as native Maven artifacts. OAuth only.

• import_deck_artifactOAuth only

  Import a finished slide deck as a self-contained HTML document, published as a native Maven deck.

  Try: “Import this deck so I can present it from Maven.”

• get_deck_artifactOAuth only

  Fetch a deck’s stored HTML document for editing and re-import.

  Try: “Get the source of the QBR deck so I can update it.”

• import_sheet_artifactOAuth only

  Import a finished .xlsx workbook as a native Maven sheet with in-app preview and download.

  Try: “Import this budget workbook into Maven.”

• get_sheet_artifactOAuth only

  Fetch a sheet’s stored workbook for regeneration and re-import.

  Try: “Download the workbook behind the media plan sheet.”