generated: '2026-08-13'
method: searched
status: published
source: https://clarisights.com/marketing-data-mcp
sources:
- https://clarisights.com/marketing-data-mcp
- https://clarisights.com/llms.txt
- https://github.com/clarisights/claude-plugins
server:
  name: Clarisights Marketing Data MCP
  transport: http
  url_pattern: https://clarisights.com/dashboard/mcp-settings
  note: 'Remote, workspace-scoped MCP server. There is no single public endpoint: each workspace gets
    its own connection URL. Setup is three published steps — (1) a workspace admin flips the global toggle
    at Settings -> Workspace -> AI access, (2) a signed-in user opens their MCP settings page and copies
    the workspace connection URL, (3) the user pastes that URL into Claude, ChatGPT, Cursor, Gemini or
    any MCP-compatible client and signs in with their Clarisights credentials. No separate API token is
    issued and no parallel permission tree exists; the same workspace ACL applies ("the AI sees what you
    see").'
deployment:
  mode: none
  endpoint: https://clarisights.com/dashboard/mcp-settings
  verified: probed
  probe: dead
  note: the endpoint this manifest claimed did not answer; recorded as none rather than deleted so the
    claim stays auditable
  checked: '2026-08-12'
  source: catalog MCP census
auth:
  type: workspace-login
  scoped: workspace
  admin_gate: Settings -> Workspace -> AI access (global on/off per workspace)
  separate_token: false
  acl: same as the signed-in user's Clarisights workspace ACL (accounts, regions, brands, reports)
  data_retention: The MCP is documented as an interface, not a store — it routes queries to Clarisights
    and returns results; documented GDPR-compliant with no user-level data stored by the MCP layer.
capabilities:
  read_only: true
  write_actions: false
  write_note: Documented as read-only. The provider's own comparison table explicitly marks "Execute campaigns
    — pause, scale, change budgets" as an ad-platform MCP capability that Clarisights does NOT offer.
  history_window: typically two full years per workspace
  semantic_layer: Queries run against one governed semantic layer with 50+ connectors joined at deepest
    granularity, expressed in the team's own KPI definitions (blended CAC, contribution-margin ROAS, MER,
    target-CPI).
  clients_named:
  - Claude
  - ChatGPT
  - Cursor
  - Gemini
tools:
- name: list_data_sources
  category: discover
  description: List available channels (e.g. facebook, adwords, tiktok) and segments — saved cross-channel
    views with pre-applied filters — in the workspace.
- name: list_metrics
  category: discover
  description: List the workspace's metrics/KPIs (CPA, ROAS, blended CAC), ranked by the team's actual
    usage, custom metrics surfaced first. Returns a metric `key` plus `display_name` and a `description`.
  parameters:
  - name: search
    description: Find metrics by name instead of paging through hundreds of results.
  - name: include_system_metrics
    description: Defaults to custom metrics only; set true (with a specific `channel`) to include single-channel
      system metrics.
  - name: channel
    description: Restrict system metrics to one channel.
- name: list_dimensions
  category: discover
  description: List breakdowns the agent can group by — channel, country, campaign, creative concept,
    audience. Custom dimensions (source "custom") work across channels; system dimensions are channel-specific.
  parameters:
  - name: search
    description: Find dimensions by name.
- name: query_data
  category: query
  description: Run aggregate, group, or trend queries with filters, date ranges and currency conversion.
    Returns clean CSV the agent reasons over; ratio metrics (CPA, ROAS, CTR) come back pre-calculated
    and "-" means no data rather than zero.
  parameters:
  - name: channels
    description: Array of channel names from list_data_sources, or a segment name (e.g. "segment_108").
  - name: metrics
    description: Array of metric keys (the `key` field, not display_name).
  - name: dimensions
    description: Array of dimension keys for breakdowns; omit for aggregate queries.
  - name: call_type
    description: '"aggregate" for totals, "groups" for dimension breakdowns, "trends" for time-series.'
  - name: date_start
    description: YYYY-MM-DD.
  - name: date_end
    description: YYYY-MM-DD.
  - name: filters
    description: 'OR-of-AND form [[{"dimension": "...", "operator": "$in", "value": ["..."]}]]. Operators:
      $in, $nin, $i_contains, $i_not_contains, $eq, $gt, $lt.'
  - name: sort_columns
    description: '[{"metric_key": -1}] for descending.'
  - name: page_size
    description: Default 50, max 1000.
  - name: to_currency
    description: Convert monetary metrics to a currency (e.g. "EUR", "USD").
- name: list_reports
  category: navigate
  description: Every Clarisights report the signed-in user has access to, surfaced to the agent for context.
- name: get_report
  category: navigate
  description: Tab-level and widget-level structure of a report, so the agent can summarize or recreate
    it.
tool_schema_provenance: 'Tool names come from the provider''s Marketing Data MCP page and llms.txt. Parameter
  lists for list_metrics, list_dimensions and query_data are transcribed from the provider''s own published
  Agent Skill (github.com/clarisights/claude-plugins, "Using the Clarisights MCP" section) — they are
  the provider''s documented call contract, not an authenticated tools/list introspection. Live tools/list
  introspection was not possible: the endpoint is workspace-scoped and requires an authenticated session,
  so no anonymous JSON-RPC probe exists to run.'
skills:
- skills/_index.yml