Tool reference

Every tool the Kula Intelligence MCP server exposes, with read-only/destructive flags and the minimum permission level.

This is the catalogue of tools the MCP server exposes. Each is annotated Read (read-only) or Write / Destructive, and the minimum permission level it needs.

The server is the source of truth — ask a connected client to list its tools to see exactly what's advertised to you. Tools you call without a high enough permission level return a clear 403 and reveal nothing.

Conventions. Read-only tools carry readOnlyHint. Tools that modify or delete data carry destructiveHint. Creating or saving workspace data (saved views, notes, skills, artifacts) requires at least the operations level; deleting them and other destructive/admin actions require admin. No tool changes member, booking or payment records at any level — those are read-only to the assistant. The exception is the studio's shared memory — the assistant's own scratchpad, not business data — where creating and editing notes is available at any level (see Memory).

Querying your data

Tool	Kind	Min level	What it does
list_tables	Read	analytics	List the tables available in your studio's database.
get_table_schema	Read	analytics	Column definitions for a table.
execute_query	Read	analytics	Run a read-only SELECT/WITH query. No writes. Schema: schema reference.
build_query	Read	analytics	Help compose a parameterised SELECT. Schema: schema reference.
entity_lookup	Read	analytics	Resolve a name or email to a canonical entity id (member, location, org).
get_member_payments	Read	operations	A member's sales/payment history (honours scoping).
get_member_plan_status	Read	operations	A member's plan/membership status.

Business context & catalogue

Tool	Kind	Min level	What it does
get_semantic_catalogue	Read	analytics	The data dictionary plus operator-authored business context for this studio.
get_business_context	Read	analytics	The studio's profile, classes, demographics, goals, brand voice, and pricing.

Graph & signals

The relationship graph and at-risk queue. See Graph & relationship tools for the full model (edges, resilience, vulnerability quadrants, signals) and per-tool detail.

Summary rollups:

Tool	Kind	Min level	What it does
get_member_affinity	Read	operations	A member's primary coach, top-3 staff, and affinity concentration.
get_staff_concentration	Read	operations	Per-staff routing value and defection risk.
get_open_signals	Read	operations	The open at-risk queue (drift, freq_softening, pause_drift, at_risk_threshold, missed_second_visit, milestone).
get_cac_by_cohort	Read	analytics	Customer-acquisition cost broken down by cohort.

Relationship context — each takes a required purpose that scopes the response (analysis purposes anonymise; service-delivery purposes withhold the resilience analytics). Denials return an empty result, not an error:

Tool	Kind	Min level	What it does
get_member_context	Read	operations	One member's resilience, vulnerability quadrant and affinity routing.
get_staff_context	Read	operations	A staff member's profile, concentration and locations.
get_location_context	Read	operations	A location's operational resilience and aggregate quadrant spread.
list_quadrant_members	Read	operations	The members in a vulnerability quadrant at a location, most-vulnerable first.
simulate_departure	Read	operations	Who's at risk if a staff member left a location.
get_entity_edges	Read	operations	The graph edges around a member/staff/location.

Views

Tool	Kind	Min level	What it does
list_views	Read	analytics	List saved views.
describe_view	Read	analytics	A view's SQL and metadata.
create_view	Write	operations	Create a named, saved view.
drop_view	Destructive	admin	Delete a view.

Skills

Tool	Kind	Min level	What it does
list_skills	Read	analytics	List available skills (built-in + your own).
get_skill	Read	analytics	Fetch a skill's full definition (may be operator-authored).
list_skill_files	Read	analytics	List a skill's bundled files.
get_skill_file	Read	analytics	Fetch one bundled skill file.
create_skill	Write	operations	Save a custom skill (lands as a draft for review).
deprecate_skill	Destructive	admin	Mark a skill deprecated.

Memory (shared, per-studio)

Tool	Kind	Min level	What it does
memory_view	Read	any	Read a note from the studio's shared memory.
memory_create	Write	any	Create a memory note.
memory_edit	Write	any	Edit a memory note.
memory_rename	Write	operations	Rename a memory note.
memory_delete	Destructive	admin	Delete a memory note.

This is the studio's shared memory, stored in this connector's database — separate from the AI client's own conversation memory. It's the assistant's scratchpad, not studio business data, so storing a note — memory_create / memory_edit — works at any level, including a read-only login that can't change anything else. Renaming needs operations; deleting (irreversible) needs admin.

Knowledge wiki

Tool	Kind	Min level	What it does
search_knowledge	Read	operations	Search the studio's knowledge notes.
get_knowledge	Read	operations	Fetch a knowledge note.
add_note	Write	operations	Add a knowledge note.
link_knowledge	Write	operations	Link knowledge notes/entities.

Semantic search

Tool	Kind	Min level	What it does
get_embeddings_status	Read	analytics	Whether semantic search is enabled for the studio.
set_embeddings_enabled	Write	admin	Turn semantic search on/off (opt-in).

Artifacts

Tool	Kind	Min level	What it does
create_artifact	Write	operations	Save a report/export/summary that persists beyond the chat.

Ignition (the diagnostic engine)

Studio Ignition is a separate paid marketplace add-on — these tools are available only on studios that have added it. See Studio Ignition.

Tool	Kind	Min level	What it does
get_ignition_status	Read	operations	Current diagnostic state.
get_ignition_intake	Read	operations	The intake form schema.
save_ignition_intake	Write	operations	Submit intake answers.
run_ignition	Destructive	admin	Run a diagnostic stage.
close_ignition_gate	Write	admin	Finalise a diagnostic stage gate.

Evals

Tool	Kind	Min level	What it does
list_eval_cases	Read	operations	List a skill's test cases.
get_eval_summary	Read	operations	Aggregate eval metrics.
add_eval_case	Write	operations	Add a test case.
record_eval_result	Write	operations	Record one eval result.

Ingest audit (raw records)

Tool	Kind	Min level	What it does
list_raw_records	Read	admin	Query the raw ingest log.
get_raw_record	Read	admin	Fetch one raw ingest payload.
record_raw	Write	admin	Write a raw ingest payload (ingest tooling).

Logs & feedback

Tool	Kind	Min level	What it does
get_error_summary	Read	operations	Aggregated errors from ingest jobs — the connector health view.
get_mcp_usage_stats	Read	operations	Per-tool usage metrics for the studio.
submit_feedback	Write	analytics	Submit feedback on a tool result (append-only).

What you won't find

No tools move money, issue refunds, generate media, or read across studios. Those capabilities are deliberately absent — see Security & data handling.