MCP tools reference

This page is generated from the MCP server's tool registry. It lists the tools returned by tools/list to a customer workspace. Descriptions and parameters are the ones the server advertises; only the first sentence of each description is shown.

The list is the same for every role. Permissions are enforced when a tool is called: a tool that changes the context layer queues the change for approval if your role may only propose edits, and SQL tools apply your role's schema, column and row permissions.

Every tool also accepts two optional parameters that are not repeated below: client_id (the workspace to act on, needed only if you belong to more than one) and user_intent (a short statement of what the user asked for, recorded for audit).

analyze_term_patterns

Resolve ambiguous business terms, abbreviations, and domain jargon by looking up how they're used in production SQL queries and certified metric definitions.

Parameter	Type	Required	Description
terms	array of string	yes	Business terms to analyze (e.g., ['d0', 'conversion', 'TS2P'])

browse_warehouse

Browse the LIVE connected data warehouse at dataset/schema/table granularity, including tables NOT yet in the context layer.

Parameter	Type	Required	Description
level	string (datasets, tables, columns)	yes	Granularity to browse. One of: datasets | tables | columns.
dataset_id	string	no	Schema/dataset filter (tables level) or target (columns).
table_name	string	no	Table to inspect at the columns level. May be a fully-qualified dataset.table — it's split for you.
table_id	string	no	Alias for table_name at the columns level: pass the table_id returned by level=tables verbatim.

call_api

Gateway to rarely-used / power-user operations that are not individual MCP tools: admin ops (view_as_role, view_as_person, manage_people, manage_members to list, invite, re-role or remove workspace members, upsert/delete_app_dataset, manage_vdm_access to create VDM schemas and grant roles on them), manage_sharing to share a data app or chat inside the workspace and answer access requests, context-table management (manage_context_tables: editors poll status, add/remove are owner-only), audit-log queries, restoring deleted context entities, triggering a context build (run_builder), area/skill/workflow housekeeping, data-app action registry, missions, join diagnostics, eval-set management, Eval Lab runs (run_evals for the whole set, one area or single questions; get_eval_run for the verdicts), and the context-edit approval queue (review_context_queue to list pending edits, review_pending_edit to approve/reject them by id or in bulk), and this workspace's credit usage this billing cycle (get_credit_usage).

Parameter	Type	Required	Description
operation	string	yes	Operation name — see list_operations.
params	object	no	Arguments for the operation, per its schema in list_operations.

call_procedure

Invoke a pre-approved warehouse stored procedure by name (Redshift only).

Parameter	Type	Required	Description
proc	string	yes	Schema-qualified procedure name, e.g. 'analytics.sp_refresh_redshift_queries'. Must be allowlisted.
args	array of any	no	Optional positional scalar args (str/int/float/bool/null), bound as parameters — never string-interpolated.

create_area

Create a domain area (maps 1:1 with metric categories).

Parameter	Type	Required	Description
name	string	yes	Area display name (e.g. 'Revenue', 'Engagement'). Also becomes the metric category.
description	string	no	Short description of this domain area.

describe_action

Inspect a single integration action — returns its agent-facing summary, side-effect classification, typed JSON Schema for the params argument to execute_action, and any registered usage examples.

Parameter	Type	Required	Description
integration	string	yes	Integration name (e.g. 'slack', 'notion').
action_type	string	yes	Action type within the integration (e.g. 'send_message', 'create-page').

dry_run_sql

Estimate a SQL query's scan WITHOUT executing it, on engines that support a dry run (BigQuery, Snowflake, Redshift); other engines return skipped.

Parameter	Type	Required	Description
sql	string	yes	The SQL query to estimate cost for
metric_id	string	no	Optional UUID of the certified metric this query is derived from. Used by certified-query enforcement; see execute_sql.metric_id.

execute_action

Execute an integration action.

Parameter	Type	Required	Description
integration	string	yes	Integration name (e.g. 'slack', 'notion').
action_type	string	yes	Action type within the integration (e.g. 'send_message', 'create-page').
params	object	yes	Action input parameters. Must conform to the typed schema returned by describe_action(integration, action_type) — unknown keys are rejected at the MCP boundary.

execute_sql

Execute a SQL query on the user's data warehouse and return results.

Parameter	Type	Required	Description
sql	string	yes	The SQL query to execute on the warehouse
is_intermediate	boolean	no	Set true when this is an intermediate lookup/discovery step (e.g., finding a name, checking values) — NOT the final answer. Intermediate results won't be validated against the question.
skip_dry_run	boolean	no	Skip the automatic pre-flight column validation (EXPLAIN runs before execution by default and catches undefined/misspelled columns without scanning data). Leave false unless you've already validated this exact SQL or the pre-flight is producing a false rejection on a valid query.
rules_token	string or array of string	no	The rules_token from the final page of your get_usage_rules read (one, or a list from several reads). Each queried table that has rules must be covered by one, or the result carries a non-fatal rules_incomplete.
max_rows	integer	no	Cap on rows returned in the response (default: 100, max: 500). When the warehouse returns more rows than this, the response is truncated to max_rows and truncated: true is set — bump max_rows up to 500 if you need more.
no_limit	boolean	no	Skip the auto-injected SQL LIMIT clause (use when you have your own LIMIT, or for aggregations). Independent of max_rows: the response is still capped at max_rows regardless. To fetch more rows, raise max_rows (up to 500), don't just set no_limit.
metric_id	string	no	Optional UUID of the certified metric this query is derived from. Required when the client has certified-query enforcement enabled and the caller's role lacks sql_full_access — the SQL must be structurally equivalent to that metric's sql_template (predicate-only deltas allowed).
label	string	no	Short human title for this result set (e.g. 'Monthly revenue'), shown as its section header in the app. Give one whenever a turn runs multiple non-intermediate queries.

execute_workflow

Trigger a workflow execution.

Parameter	Type	Required	Description
workflow_id	string	yes	UUID of the workflow to execute
input_data	object	no	Input parameters for the workflow (e.g. user_id, date_range). Check trigger node's input_schema for required fields.
trigger_type	string (test, manual)	no	Trigger type (default 'test', or 'manual' when send_slack is true; a send_slack run can't be 'test')
send_slack	boolean	no	Default false. Set true ONLY when the user has explicitly asked for this run to post its Slack messages for real (e.g. a scheduled job delivering a digest). Enables the workflow's Slack send_message nodes and nothing else: Slack approval cards, email, HubSpot, Attio, webhooks and every other integration stay skipped, and so does a Slack node marked schedule_only. Must be a JSON boolean; anything else is refused. The reply's integrations_enabled / integrations_skipped say per node what will happen.
from_scratch	boolean	no	Re-run VDM (structured-map) nodes from scratch: ignore the per-row LLM cache and re-enrich every row. Default false — an ordinary run reuses cached per-row results even after a prompt edit. Use this to deliberately re-classify the whole model (e.g. after changing a prompt).
max_rows	integer	no	Rows processed per source. Omit for the default capped test run (a data node with materialize that sets no max_rows of its own builds in full instead — up to 50,000 rows — when the workflow has no per-row LLM fan-out, since a capped read could never refresh the model); pass 0 to process ALL records (a full run: up to 50,000 rows per source when the run executes on the AWS Batch tier; 5,000 when it executes on the app server — environments without a Batch tier, workflows with agent/checklist nodes or a skill node that renders outside a skill bundle (they need the context layer), a failed Batch submit, or local dev). The run result reports if/where it was capped — do a test run first, then re-run with max_rows=0 once it looks right.
bypass_row_cap	boolean	no	Only meaningful together with max_rows=0, and only for a run that executes on the app server rather than the AWS Batch tier (environments without a Batch tier, workflows with agent/checklist nodes or a skill node that renders outside a skill bundle (they need the context layer), a failed Batch submit, or local dev): there a full run is capped at 5,000 rows so an unbounded per-row LLM fan-out can't wedge the shared workers, and this raises it to 50,000 for a pure-SQL build — e.g. a data node with materialize set building a mart straight from the warehouse. The server ENFORCES the precondition: if the workflow contains any per-row LLM fan-out (a loop node with mode each_row, or a node with loop_mode each_row), the request is ignored, the run stays at 5,000, and the run's notices say so. On the AWS Batch tier a full run already reaches 50,000 rows, per-row steps included, so the flag changes nothing there — but it is harmless to pass when unsure where the run will execute.

find_join

Find how two tables can be joined, or list all tables joinable to one table.

Parameter	Type	Required	Description
from_table	string	yes	Source table (fully qualified, e.g. 'analytics.dim_user')
to_table	string	no	Optional target table — if provided, returns the join path

generate_sql

DEPRECATED — does NOT generate SQL.

Parameter	Type	Required	Description
tables	array of string	yes	List of table IDs to use in the query
question	string	yes	The natural language question (echoed back; no longer used to generate SQL)

get_area

Get a hydrated domain area bundle: area metadata plus all tagged tables, metrics in this category, scoped docs (filenames — fetch contents with get_document), usage-rule pointers, and business terms.

Parameter	Type	Required	Description
area_id	string	yes	Area ID (slug, e.g. 'revenue', 'engagement')

get_business_context

Get the business primer for this client's data warehouse.

Parameter	Type	Required	Description
area_id	string	no	Optional area ID. When set, returns area-specific context (tables, metrics, usage-rule pointers, docs, and the global rules in full) instead of the global overview. Full rule text: get_usage_rules(area_id=...).

get_document

Read the full markdown of an area-scoped Domain Context Pack doc.

Parameter	Type	Required	Description
filename	string	yes	Document filename as returned in get_area's docs[].filename (e.g. 'area_conversion.md').
section	string	no	Optional ## heading to return on its own. Omit for the whole document; the response always lists the available headings under sections.

get_pipeline_health

Health snapshot of every pipeline Rig runs for this client: data-ingestion source syncs (what came in overnight, rows loaded, errors), post-sync dbt model builds, the context engine's scheduled gardener/builder, and scheduled workflows.

Parameter	Type	Required	Description
all_clients	boolean	no	Rig platform admins only: roll up health across ALL app.rig.so clients instead of just this one.

get_pipeline_runs

Recent pipeline run history for this client — ingestion syncs, workflow executions, dbt model builds, and context builder/gardener runs, newest first, with status, trigger, duration and error.

Parameter	Type	Required	Description
days	integer	no	Lookback window in days (default 7, max 90).
limit	integer	no	Max runs to return (default 100, max 500).

get_playbook

Load a specific playbook by name.

Parameter	Type	Required	Description
name	string	yes	Playbook name (e.g. 'workflows', 'sql_authoring')
section	string	no	Optional ## heading to return on its own (e.g. 'Dashboard output'). Omit for the whole playbook; the response always lists the available headings under sections.

get_shared_state

Read state that people curated in data apps and shared across the platform — review statuses, overrides, classifications, notes — keyed by namespace: 'org' / 'org:' (tenant-wide), 'area:' (an area), 'entity::' (one business entity, e.g. entity:account:acme).

Parameter	Type	Required	Description
namespace	string	no	e.g. 'org', 'area:finance', 'entity:account:acme'. Omit to list namespaces.
prefix	string	no	Only keys starting with this.
since	string	no	Change-log cursor; returns changes after it.

get_table_columns

Get the full schema for one or more warehouse tables: all columns with their data types, descriptions, and sample values.

Parameter	Type	Required	Description
table_id	string or array of string	yes	Fully-qualified table name(s) — pass a single string, or a list to fetch multiple in one call.
column_filter	array of string or string	no	Columns to return — a list of names/patterns (a column is kept if it matches ANY), or a single one. Each entry is an exact name, a fuzzy substring (e.g. 'revenue') or a LIKE pattern (e.g. '%_id'). Put every column you need in one list rather than calling once per column.
include_all	boolean	no	Return all columns even if >60. Default false.
include_empty	boolean	no	Include all-NULL / dead columns (e.g. dlt _v variant fossils) that are hidden by default. Default false.

get_usage_rules

Get non-obvious routing rules, at table, area, or global (context-wide) scope.

Parameter	Type	Required	Description
table_ids	array of string or string	no	Table IDs to get rules for — pass a JSON array (e.g. ['analytics.fct_orders', 'analytics.dim_customers']), a single table string, or a comma-separated string.
area_id	string	no	Filter to rules scoped to this area, plus the rules on its tagged tables.
cursor	string	no	Continue a paged read — only ever the quoted token inside a response's next_page (cursor='…'), with the same other arguments.
scope	string (global)	no	scope='global' alone fetches only global (context-wide) rules — no table_ids needed. Combined with table_ids/area_id, the full scoped read runs.
include_global	boolean	no	Include the tenant-wide global rules (default true). Pass false when you already have them from get_business_context or an earlier call.
include_area_rules	boolean	no	With table_ids: also include the area-scoped rules of every area those tables are tagged to (default true). Pass false for just the tables' own rules.

inspect_table

Get details about a specific warehouse table.

Parameter	Type	Required	Description
table_id	string	yes	Fully-qualified table name (e.g. 'analytics.dim_user') or partial name for what='resolve'
what	string (resolve, samples, lineage, upstream, patterns)	yes	What to inspect about the table
depth	integer	no	For what='upstream' or 'lineage': depth 1-3
column_name	string	no	For what='lineage'/'upstream': trace ONE column through synced warehouse view definitions (source object/field -> landed column -> view columns, with parsed/inferred marks) via the response's view_lineage
column_names	array of string	no	For what='samples': filter to these columns

inspect_template

Read a node's template — whole, in part, or as a map.

Parameter	Type	Required	Description
workflow_id	string	yes	UUID of the workflow — or whatever the user pasted for the app: a data-apps URL (?workflow=...) or a public embed link / share id (/skills/embed/), both resolve to it
node_id	string	yes	Node ID to inspect
target	string	no	Read one element: '#id', 'tag#id', or '[attr=value]'.
region	string	no	Read one named region by id.
inner	boolean	no	With target/region: return the contents without the wrapper element or markers.
outline	boolean	no	Force the outline (regions + addressable ids) regardless of size.
versions	boolean	no	Return this node's template version history instead of its content.
full	boolean	no	Return the entire body even when it is large.
max_depth	integer	no	Outline nesting depth (default 8). Raise it if elements_omitted reports ids deeper than the cut.

inspect_workflow_run

Inspect a workflow run.

Parameter	Type	Required	Description
run_id	string	yes	UUID of the workflow run
what	string (detail, errors, node_output)	yes
node_id	string	no	Required for what='node_output'
include_outputs	boolean	no	For what='detail': embed each node's full output payload. Default false (summary-only).
row_filter	string	no	For what='node_output': filter rows where a column equals a value, formatted as 'column=value' (e.g. 'status=open'). Default returns full output.
max_chars	integer	no	For what='node_output': character cap on the returned payload. Default 4000.

invoke_workflow_action

Operate one of a data app's declared actions (a button) by id — the same effect as a user clicking it.

Parameter	Type	Required	Description
workflow_id	string	yes	UUID of the workflow — or whatever the user pasted for the app: a data-apps URL (?workflow=...) or a public embed link / share id (/skills/embed/), both resolve to it
action_id	string	yes	Action id from list_workflow_actions
params	object	no	Per-invocation values merged over the action's declared defaults (e.g. the account_id this decision applies to).

list_actions

List integration actions available to the current user.

Parameter	Type	Required	Description
integration	string	no	Optional filter — pass an integration name (e.g. 'slack', 'notion', 'hubspot', 'chargebee') to list only that integration's actions. Omit to return all integrations.

list_playbooks

List the task-specific playbooks available for this MCP server.

list_skills

List skills visible to the current user, newest-first.

Parameter	Type	Required	Description
owner	string (me, org)	no	Filter by ownership.
tag	string	no	Filter to skills tagged with this string.
limit	integer	no	Max results (default 50).

list_workflow_actions

List the interactive actions (buttons) a data app exposes — e.g. an approve/reject or a custom action that records a decision to a VDM or fires an integration.

Parameter	Type	Required	Description
workflow_id	string	yes	UUID of the workflow — or whatever the user pasted for the app: a data-apps URL (?workflow=...) or a public embed link / share id (/skills/embed/), both resolve to it

load_skill

Return the full markdown of a specific skill, by id.

Parameter	Type	Required	Description
skill_id	string	yes	Skill UUID from search_skills / list_skills.
version	integer	no	Optional version number; defaults to the current published version.
agent_key	string	no	Optional deterministic agent key for Smart Skills attribution, e.g. chat:semantic or workflow::.
trace_id	string	no	Optional trace id to join skill usage to cost/outcome telemetry.
run_id	string	no	Optional chat/workflow run id to join skill usage to outcomes.
session_id	string	no	Optional external agent session id.

manage_business_term

Create, update, or delete a business term (domain vocabulary entry). action='create' (term_data); action='update' (term_id + term_data); action='delete' (term_id).

Parameter	Type	Required	Description
action	string (create, update, delete)	yes
term_id	string	no	Required for update/delete
term_data	object	no	Required for create/update. Keys: 'term', 'definition', 'aliases', 'category', 'area_id', plus 'table_refs' / 'metric_refs' — lists of table ids / metric names this term maps to. THIS IS THE ONLY WAY TO LINK A METRIC OR TABLE TO A TERM: metrics have no business_term_refs field, so the relation is written here in both directions. Omitting a refs key leaves it untouched; passing [] clears it.

manage_ingest

Trigger and monitor data-connection syncs (rig_ingest).

Parameter	Type	Required	Description
action	string (status, list_connectors, fabric_coverage, trigger, set_streams, route_streams, set_schedule, pause, resume, cancel)	no	'status' (default) reads sync state; 'list_connectors' lists addable connectors; 'trigger' starts a sync; 'set_streams' turns streams on/off; 'set_schedule' sets the sync frequency; 'pause' / 'resume' switch the source off/on; 'cancel' stops a running sync; 'fabric_coverage' reads what Fabric Link already lands; 'route_streams' moves streams between Fabric Link and Rig.
source_name	string	no	Source to act on or report on (e.g. 'fireflies'). Required for every action but 'status' / 'fabric_coverage' (omit = all sources) and 'list_connectors'.
full_resync	boolean	no	action='trigger' only: re-pull ALL history instead of an ordinary sync. Owner only; confirm with the user first.
enable	array of string	no	action='set_streams': stream names to turn on (e.g. ['adset_daily']).
disable	array of string	no	action='set_streams': stream names to turn off.
to	string (fabric, rig)	no	route_streams only: 'fabric' = let Fabric Link supply them, 'rig' = sync currently skipped/excluded streams with Rig again.
streams	array of string	no	route_streams only: stream names (e.g. ['accounts']). Omit with to='fabric' for every covered stream.
frequency	string (nightly, hourly, manual, custom)	no	action='set_schedule': 'nightly' / 'hourly' (UTC), 'manual' (no automatic syncs), or 'custom' (with cron).
cron	string	no	frequency='custom' only: 5-field cron (e.g. '30 6 * * 1-5') or '@workday N HH:MM'.
timezone	string	no	frequency='custom' only: IANA zone the schedule runs in (omitted = the source's current zone).
search	string	no	action='list_connectors': filter by name, display name or category (e.g. 'crm', 'pendo').

manage_metric

Create, update, or delete a certified metric. action='create' (metric_data); action='update' (metric_id + metric_data); action='delete' (metric_id).

Parameter	Type	Required	Description
action	string (create, update, delete)	yes
metric_id	string	no	Required for update/delete
metric_data	object	no	Required for create/update — full metric spec (readable_name, description, sql, primary_table, entity, measure, grain, derived, etc.). When 'sql' is supplied it is persisted verbatim — no LLM rewrite. When 'sql' is omitted on create, an LLM drafts SQL from description; on update, an LLM re-drafts SQL only if 'description' is supplied without an explicit 'sql'. 'category' is the metric's primary/home area (the area display name). 'area_ids' is an optional list of additional area slugs the metric also belongs to (secondary/tertiary areas) — it makes the metric surface in those areas' bundles without changing its home area. 'dependent_tables' (alias 'table_ids') is the metric's source tables. To link a metric to a business term use manage_business_term's 'metric_refs' — metrics have no business_term_refs field.

manage_usage_rule

Create, update, or delete a usage rule (non-obvious routing rule — wrong-table traps, required filters, grain surprises).

Parameter	Type	Required	Description
action	string (create, update, delete)	yes
rule_id	string	no	Required for update/delete
rule_data	object	no	Required for create; optional on update (only 'rule_id' is required there — supplied keys are applied as edits; 'table_id' is ignored on update). Create keys: 'rule_text' (the rule body the agent should follow), plus one of three scopes: 'table_id' (single canonical id, not a list — use search_tables to resolve) for a table-scoped rule; 'scope'='area' PLUS 'area_id' for an area-scoped rule (a bare 'area_id' is not enough — the scope must be explicit); or 'scope'='global' for a tenant-wide rule (applies everywhere, no table_id/area_id needed). Optional: 'rule_type' (usage_note|required_filter|wrong_table_trap, default usage_note). On update, 'scope'/'area_id' RE-SCOPE the rule (scope='global' makes an area rule tenant-wide; an area_id on a global rule confines it to that area). Do NOT pass 'tables', 'title', 'description', 'user_intent' — those are not part of the schema and the call will be rejected.
table_id	string	no	Top-level convenience form of rule_data.table_id.
rule_text	string	no	Top-level convenience form of rule_data.rule_text.
content	string	no	Alias for rule_text.
rule_type	string	no	Top-level convenience form of rule_data.rule_type (usage_note|required_filter|wrong_table_trap).
scope	string	no	Rule scope: table (default) | area | global. area/global apply beyond the analysis at hand — use them only when the user explicitly asked to record the rule, or confirmed one you proposed; a convention merely stated as an FYI is not an ask. On action='update' this RE-SCOPES the rule (e.g. scope='global' moves an area rule to tenant-wide). Top-level convenience form of rule_data.scope.
area_id	string	no	Area for scope='area'. On action='update', setting it on a global rule re-scopes that rule to the area. To go the other way send scope='global' — a null area_id is only honoured inside rule_data. Top-level convenience form of rule_data.area_id.

manage_workflow

Workflow definition CRUD. action='create' (workflow_data), action='update' (workflow_id + updates: nodes merges by id and never drops a node you leave out; remove_node_ids is the only way to delete a node — an unrecognised update field is an error, not ignored), action='get' (workflow_id), action='list' (optional is_active + search filters), action='delete' (workflow_id) — soft delete, same as the app UI's delete: the workflow disappears from lists and stops running, but its materialized derived.* tables are retained (the response lists any; drop one with manage_derived_model action='delete').

Parameter	Type	Required	Description
action	string (create, update, get, list, delete, set_schedule, clear_schedule)	yes
workflow_id	string	no	Required for get/update/delete/set_schedule/clear_schedule. A workflow UUID, or anything the user pasted for the app: a data-apps URL (?workflow=...) or a public embed link / share id (/skills/embed/) — both resolve to the workflow.
schedule	object	no	Required for action='set_schedule'. Writes the workflow_schedules row the scheduler actually reads.
workflow_data	object	no	Required for create
updates	object	no	Required for update — partial node/config changes. Also accepts the /data-apps organisation fields: category (the folder) and lifecycle_state ('draft' | 'published', a curation label that never gates visibility or execution). An update touching ONLY those is filing, not authoring: it takes no version snapshot and does not bump the version. Accepted fields: name, description, nodes, remove_node_ids, context_layer, category, lifecycle_state, plus expected_revision — anything else is rejected by name, as are fields placed beside updates instead of inside it.
is_active	boolean	no	Optional filter for action='list'
search	string	no	Optional name/description substring filter for action='list'
authoring_notes	object	no	REQUIRED when creating or updating a skill node. Capture the WHY behind this skill, drawn from your own conversation context — not a transcript dump. Three short paragraphs (each 1–4 sentences): user_question — the original ask, as close to verbatim as you can reproduce exploration — what you checked: tables searched, columns inspected, SQL drafts tried, what worked vs didn't design_decisions — why these tables/joins/templates were chosen over alternatives Concrete > vague. Mention specific table/column names, SQL snippets, and rejected paths. Optional for non-skill workflows but encouraged.

prepare_skill_run

Prepare a skill workflow for execution by an external agent.

Parameter	Type	Required	Description
workflow_id	string	yes	UUID of the skill workflow
force_revalidate	boolean	no	If true, re-validate even templates verified within the last 24h. Default false.

propose_chart

Propose a chart to visualize your LAST tabular result — an execute_sql query OR a workflow run's data read via inspect_workflow_run(what='node_output').

Parameter	Type	Required	Description
data_ref	string	no	Which result set to chart: a data_ref echoed by an earlier execute_sql result (e.g. 't3.q1'), including from previous turns. Omit to chart your latest result this turn.
chart_spec	object	yes

publish_template

Promote a staged draft to live, or roll back to an earlier version.

Parameter	Type	Required	Description
workflow_id	string	yes	UUID of the workflow — or whatever the user pasted for the app: a data-apps URL (?workflow=...) or a public embed link / share id (/skills/embed/), both resolve to it
node_id	string	yes	Node ID with a staged draft, or whose template to restore
verify_render	boolean	no	Load the dashboard headlessly before it goes live and refuse the publish if it throws or renders blank (default true).
restore_version	integer	no	Roll this node's template back to the given workflow version instead of publishing the draft. Omit it to publish the draft (0 is treated as omitted).

refresh_table_schema

Re-scan the named tables against the LIVE warehouse schema and pull in any columns Rig hasn't indexed yet.

Parameter	Type	Required	Description
table_ids	array of string	yes	Schema-qualified table ids to re-scan, e.g. ['dbt_staging.stg_hubspot_companies']. Use the table_id values from search_tables / browse_warehouse.

regenerate_context

Regen of derived context state: rebuilds the DuckDB cache from the fde_* tables, recomputes embeddings for new/changed rows (existing embeddings hit cache), atomically swaps in the new file, exports parquet, and invalidates in-memory caches.

save_skill_run

Save the result of a skill execution run by an external agent (e.g. Claude Code / Cowork).

Parameter	Type	Required	Description
workflow_id	string	yes	UUID of the skill workflow this run belongs to
title	string	no	Short title for the run (e.g. 'SME Churn Deep Dive — Q1 2026')
result_md	string	yes	Markdown output of the skill investigation
result_html	string	no	Optional HTML output (served directly at the run URL). If omitted, markdown is auto-rendered.
inputs	object	no	Input parameters that were used (for audit trail)
metadata	object	no	Extra context: tables queried, row counts, duration, etc.

search_columns

Search for columns across all warehouse tables by name or business concept.

Parameter	Type	Required	Description
query	string	no	Column name or concept to search for (e.g., 'revenue', 'user_id', 'subscription status')
table_filter	string	no	Optional: limit search to columns in this table (e.g., 'analytics.fct_orders')
q	string	no	Shorthand alias for query — prefer query.

search_metrics

Search for certified and draft business metrics (KPIs, measures, calculations) by natural language query.

Parameter	Type	Required	Description
top_k	integer	no	Maximum number of metrics to return (default: 5, max: 20)
area_id	string	no	Filter to metrics in this area's category.
metric_ids	array of string or string	no	Fetch these metrics in full (untruncated SQL and description) by metric_id, instead of searching — up to 5 per call. No query is needed.
query	string	no	The user's question to match against metrics
q	string	no	Shorthand alias for query — prefer query.

search_queries

Search for historical analyst SQL queries that match a question.

Parameter	Type	Required	Description
top_k	integer	no	Maximum number of queries to return (default: 5)
query	string	no	The question to match against historical queries
q	string	no	Shorthand alias for query — prefer query.

search_skills

Semantic search over the Skill Library — reusable markdown Claude skills (playbooks, workflows, debugging recipes) authored by you and your org.

Parameter	Type	Required	Description
query	string	yes	What you're trying to do (e.g. 'debug Redshift WLM queue saturation').
limit	integer	no	Max results (default 10, max 50).

search_tables

Search for warehouse tables by name, description, or business concept.

Parameter	Type	Required	Description
query	string	no	Search query for finding tables
top_k	integer	no	Maximum number of results to return (default: 10)
area_id	string	no	Filter to tables tagged with this area (domain).
gold_star_only	boolean	no	Restrict results to gold-star (vetted, dashboard/VDM-eligible) tables only. Gold-star tables are always ranked first regardless of this flag.
q	string	no	Shorthand alias for query — prefer query.

search_values

Find which warehouse table and column contains a specific data value.

Parameter	Type	Required	Description
value	string	yes	The value to search for (e.g., 'Balance based workers', 'succeeded')
table_filter	string	no	Optional: limit search to columns in this table

set_template_draft

Stage a node's output template in the draft slot — either as granular edits (preferred) or a whole content body.

Parameter	Type	Required	Description
workflow_id	string	yes	UUID of the workflow — or whatever the user pasted for the app: a data-apps URL (?workflow=...) or a public embed link / share id (/skills/embed/), both resolve to it
node_id	string	yes	ID of the skill or llm node to update
content	string	no	Whole template body. Mutually exclusive with edits. For dashboards, use as the data placeholder.
edits	array of object	no	Granular changes applied atomically to the current template. Mutually exclusive with content.
format	string (markdown, html_simple, dashboard)	no	Optional explicit format. Auto-detected from the resulting content if omitted.
assets	array of object	no	External script imports; server injects into in array order.
auto_publish	boolean	no	Promote the draft to live in the same call. Defaults to TRUE when edits is used, FALSE when content is used.
verify_render	boolean	no	Load the dashboard in a headless browser before it goes live and refuse the publish if it throws or renders blank (default true; adds a few seconds). Set false only when you know the check cannot run.

tag_tables_to_area

Tag one or more tables to an area.

Parameter	Type	Required	Description
area_id	string	yes	Area ID (slug)
table_ids	array of string	yes	List of table IDs to tag (e.g. ['analytics.fct_orders', 'analytics.dim_customers'])

test_sql

Test and validate a SQL query in a lightweight DuckDB sandbox before running it on the real warehouse.

Parameter	Type	Required	Description
sql	string	yes	The SQL query to test
rules_token	string or array of string	no	The rules_token from the final page of your get_usage_rules read (one, or a list from several reads). Each queried table that has rules must be covered by one, or the result carries a non-fatal rules_incomplete.

untag_table_from_area

Remove a table from an area.

Parameter	Type	Required	Description
area_id	string	yes	Area ID (slug)
table_id	string	yes	Table ID to remove

update_area

Update an area's name, description, or summary.

Parameter	Type	Required	Description
area_id	string	yes	Area ID (slug)
name	string	no	Updated display name (cascades to metric category)
description	string	no	Updated description
summary_md	string	no	Summary markdown. When the call is attributable to a signed-in person it is labelled manually edited and the gardener will not overwrite it; from an unattributed agent session it is stored as generated text, refreshable by the next automated pass.

update_description

Update the description of a warehouse table or column, or of a derived model (VDM, table_id derived.<name>) and its columns.

Parameter	Type	Required	Description
target_type	string (table, column)	yes
table_id	string	yes	Fully-qualified table name
column_name	string	no	Required if target_type='column'
description	string	yes	New description text

upload_conversation

Land an external chat transcript (today: Claude Code) as a real Rig chat.

Parameter	Type	Required	Description
dataset_id	string	no	Rig dataset to bind a hosted report to (drives which warehouse follow-up SQL runs against). report_html only — an imported transcript resolves its warehouse from the client. If omitted, falls back to the client's default.
messages	array of object	no	Pre-normalized list of message turns: [{role: 'user'|'assistant', content: str, timestamp?: ISO8601}]. Use this OR raw_events, not both.
raw_events	array of object	no	Raw Claude Code JSONL events (one parsed line per item). The tool will filter to user/assistant turns and flatten tool-use / tool-result blocks into a readable markdown summary.
title	string	no	Conversation title. Defaults to the first user turn (truncated).
source	string	no	Provenance label, defaults to 'claude-code'. Stored in metadata.source.
source_session_id	string	no	Original session id (e.g. Claude Code sessionId) — preserved for traceability.
source_metadata	object	no	Free-form provenance bag — cwd, gitBranch, entrypoint, etc. Surfaced in the provenance header and stored under metadata.source_metadata.
owner_user_id	string	no	Supabase user UUID to attribute the conversation to. Defaults to the calling token's user_id. AUTHZ: attributing to a different user requires tenant-admin (owner/admin role) in the caller's org, OR platform-admin (@rig.so email). Non-admins may only upload to themselves.
owner_email	string	no	Alternative to owner_user_id — looked up via Supabase admin API. Falls back silently to the calling token's user_id on miss (warning logged). Same authz gate as owner_user_id.
report_html	string	no	Full styled HTML document to host as a standalone, shareable report conversation (charts/inline