MCP tool reference

Last reviewed against the app on 24 September 2026

The clariBI MCP server exposes 26 tools, 2 prompts and 3 resources. tools/list returns every tool to any client; whether a call succeeds depends on the token's scopes and your plan. The live list, generated from the running server, is at /api/mcp-server/tool-catalog/.

Errors

Code	Meaning
`-32001`	Not authenticated
`-32002`	Missing scope
`-32003`	Rate limited
`-32004` / `-32005` / `-32006`	Unknown tool / resource / prompt
`-32007`	Not included in your plan
`-32602`	Invalid arguments

Public tools (no account needed)

check_pricingpublic

Plans with prices, AI credits, data source and user limits and main features.

Arguments: tier (optional: free, trial, lite, starter, professional, enterprise).

register_accountpublic

Starts signup: checks the email and organization name, emails a 6-digit code and returns a pending_id valid for 10 minutes.

Arguments: email (required), organization_name (required, 2 to 200 characters), accept_terms (required, must be true), first_name, last_name.

verify_emailpublic

Finishes signup. Returns a 1-hour access token and a long-lived MCP API key; the new organization starts on the 14-day Trial.

Arguments: pending_id, code, password (at least 8 characters), all required.

Read tools

Tool	Scope	What it returns	Arguments
`list_dashboards`	`dashboards:read`	Dashboards with status, last refresh and web URL	`limit` (1 to 100, default 20), `offset`, `search`
`get_dashboard`	`dashboards:read`	One dashboard with widget definitions and latest data	`dashboard_id`
`list_reports`	`reports:read`	Reports with status, format and download URL	`limit` (default 20), `offset`, `status`
`get_report`	`reports:read`	One report	`report_id`
`list_data_sources`	`data-sources:read`	Sources with type, status, last sync and row count	`limit` (default 50), `offset`, `source_type`
`get_data_source_schema`	`data-sources:read`	Columns of a source; status becomes `active` when processing is done	`data_source_id`
`get_usage`	`usage:read`	AI credits used, data source and user counts	none
`get_billing_status`	`billing:read`	Plan, renewal date and upgrade options	none
`list_forecasts`	`forecasts:read`	Forecasts with source, horizon, schedule and latest run	`limit`, `offset`, `is_active`
`get_forecast`	`forecasts:read`	One forecast configuration	`forecast_id`
`get_forecast_run`	`forecasts:read`	Full result of a run: projection with bands, drivers, anomalies, changepoints, summary	`forecast_id`, `run_id` (or `latest`)
`get_forecast_trust`	`forecasts:read`	How accurate past runs turned out against actuals	`forecast_id`

The four forecast tools need Starter or higher.

Write tools

run_analysisanalysis:runAI plans

Answers a question about your connected data, as in Ask clariBI. Uses AI credits. Returns the answer if it finishes within wait_seconds, otherwise a job_id.

Arguments: question (required, 3 to 1,000 characters), session_id (continue a session), wait_seconds (0 to 60, default 30).

get_analysis_statusanalysis:run

Status, and the answer when done, of a run_analysis job. Argument: job_id.

upload_data_sourcedata-sources:write

Creates a data source from a base64-encoded file, processed like a web upload. Up to 25 MB encoded (about 18 MB of file).

Arguments: name, format (csv, tsv, json, xlsx, xls, txt, pdf) and data_base64, all required; description; wait_seconds (0 to 60, default 0).

ingest_url_data_sourcedata-sources:write

Creates a data source from a file at a public http(s) URL, fetched once by the server. Up to 100 MB. Private and internal addresses are refused.

Arguments: name and url (required), format (optional, same list), description, wait_seconds.

request_oauth_integration_urldata-sources:write

Returns a URL where the user approves a Google, Meta, Jira or Confluence connection in the browser, and a handoff_id.

Arguments: provider (google, meta, jira, confluence; required), integration_type (for example ga4, google_ads, gsc, gsheets for Google; ads for Meta; default basic).

check_integration_statusdata-sources:write

Status of a handoff: pending, connected, failed or expired, and the data source ID once connected. Argument: handoff_id.

generate_reportreports:writeAI plans

Creates a report entry with status pending and returns its report_id and web URL.

Arguments: title (required, 2 to 200 characters), template_id (a template from the library), output_format (json, pdf, html, excel, csv; default pdf).

Note: at the moment this tool records the report but does not start generating it. To produce a report, use the Reports page in the app or POST /api/v1/reports/generate/ (Professional and Enterprise).

create_forecastforecasts:writeStarter+

Creates a forecast on a metric and checks the metric can be read before any credits are spent.

Required: name, source_type (data_source, report, dashboard, goal), source_id, metric_path. Optional: granularity (hourly, daily, weekly, monthly, quarterly, annual; default daily), horizon_days (buckets ahead, default 30; at most 168 hourly, 90 daily, 52 weekly, 24 monthly, 8 quarterly, 5 annual), aggregation, transform, include_correlations, include_anomalies, include_changepoints, narration_enabled, method_override, schedule_frequency (daily, weekly, monthly, manual; default monthly) and schedule day fields.

run_forecastforecasts:runStarter+

Runs a forecast now. Costs 1 to 10 credits, plus 1 or 2 for the written summary. Returns the result if done within wait_seconds (0 to 60, default 30), otherwise a run_id.

regenerate_forecast_narrativeforecasts:runStarter+

Rewrites only the AI summary of a completed run (1 to 2 credits). Arguments: forecast_id, run_id.

Billing tool

create_checkout_sessionbilling:write

Returns a Stripe Checkout URL the user opens in a browser to upgrade. No payment happens in the chat.

Arguments: tier (starter, professional, enterprise; required), billing_period (monthly or yearly; default monthly).

Plan requirements

All tools except the public ones need an organization on Trial, Starter, Professional or Enterprise. On top of that, "AI plans" tools (run_analysis, generate_report) need a plan with AI analysis (Trial, Starter, Professional, Enterprise), and forecast tools need Starter or higher.

Prompts

Prompt	What it does	Arguments
`analyze_question`	Wraps a question in an analysis prompt that lists your connected data sources.	`question`
`weekly_review`	Starts a weekly review of all connected sources: main movements, anomalies and what to look at next.	none

Both need the analysis:run scope.

Resources

URI	Content	Scope
`dashboard://<id>`	Dashboard configuration and latest data	`dashboards:read`
`report://<id>`	Generated report content	`reports:read`
`data-source://<id>/schema`	Column types and row counts	`data-sources:read`

Example: list dashboards

POST /mcp/v1/
Content-Type: application/json
Authorization: Bearer claribi_mcp_...

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": { "name": "list_dashboards", "arguments": { "limit": 5 } }
}

The result has a short text summary in content and the data in structuredContent, for example total and items with each dashboard's id, name, status and web_url.