mbql

name	mbql
description	Author and debug MBQL query bodies for the `mb` CLI — the only hand-authorable query format. Covers the JSON shape (flat numeric-id stages, options-object-second clauses, optional lib/uuid), joins and FK traversal, multi-stage pipelines, aggregation naming, the flat-vs-legacy-envelope footgun, and the print-schema → dry-run → run validation loop. Use when writing or fixing any query body — `mb query`, a card's `dataset_query`, a transform's `source.query`, or a segment/measure `definition` — or when `--dry-run`/run reports validation errors. Triggers — "write an MBQL query", "the dataset_query is wrong", "aggregate and group by", "join two tables", "month-over-month".
allowed-tools	Read, Write, Edit, Bash, AskUserQuestion

MBQL

MBQL is the query format you author by hand — it has a bundled JSON Schema, so the CLI pre-flight-validates it before sending. A native SQL query is also MBQL: its single stage is mbql.stage/native (raw SQL) instead of mbql.stage/mbql (structured), so it's pre-flight-validated the same way (its SQL string aside) — see native-sql. Only the legacy flat forms below skip validation.

Prefer a structured stage over a native SQL stage: portable across warehouse engines. Try it first; fall back to a native stage when structured MBQL can't express what you need, or when a structured body keeps failing server-side and you can't resolve it. For native SQL with parameters (template tags, field filters, snippets), load native-sql.

General flag conventions, body-input precedence, output flags, ./.scratch, and mb uuid mechanics live in core (mb skills get core).

The shape

A flat object — lib/type, a numeric database id, and an ordered stages array. No recursive source-query nesting; multi-step queries are sibling stages.

{
  "lib/type": "mbql/query",
  "database": 1,
  "stages": [
    {
      "lib/type": "mbql.stage/mbql",
      "source-table": 7,
      "aggregation": [["count", {}]],
      "breakout": [["field", { "temporal-unit": "month" }, 22]]
    }
  ]
}

• Numeric ids only. database, source-table, and field ids are integers from mb database list / mb table get <id> --include fields. (Git-sync YAML uses names like [Sample Database, PUBLIC, ORDERS]; the /api/dataset form uses numeric ids — don't mix them.)
• First stage carries source-table (a table id) or source-card (a saved card). Later stages omit both and read the previous stage's output columns by name.
• source-card references a saved card by its numeric id (from mb card list), not its string entity id; downstream fields are referenced by column name (string), not a field id.

The one rule that trips everyone: options object is second

Every clause is [op, {options}, ...args]. The options object is element 1, args follow.

["field", { "base-type": "type/Text" }, 1779]   // field id is THIRD; options may be empty {}
["count", {}]                                    // no args
["sum",   {}, ["field", {}, 42]]
["=",     {}, ["field", {}, 1779], "delivered"]
["asc",   {}, ["field", {}, 42]]

The legacy field shape ["field", id, opts] (id second) is rejected here. A slot-1 violation surfaces from --dry-run as must be the field options object / must be the clause options object at /stages/0/<verb>/<n>/1.

The same [op, {options}, …] rule holds for aggregation, breakout (a list of field refs), filters (implicitly ANDed; nest an explicit ["or", {}, …] for OR), order-by, expressions, and join conditions.

UUIDs: optional — mint only to reference a clause

lib/uuid is optional — leave it out whenever you can. Omit it and the server generates a unique one for every clause; an empty options object {} is the normal case. The more UUIDs you hand-manage the easier it is to trip the server's "all lib/uuids must be unique" check — a duplicated UUID passes pre-flight, then fails server-side.

Set an explicit lib/uuid only when you must reference a clause from elsewhere in the query — you have to know the value to point at. The case that needs it: ordering by (or otherwise reusing) an aggregation. ["aggregation", {…}, "<uuid>"]'s third arg is the string lib/uuid of the target aggregation, so give that aggregation an explicit lib/uuid and point the ref at the same string. A numeric position fails with must be the target aggregation's lib/uuid (string), not a numeric position.

"aggregation": [["count", { "lib/uuid": "AGG_UUID" }]],
"order-by":   [["desc", {}, ["aggregation", {}, "AGG_UUID"]]]

(AGG_UUID is both the aggregation's own lib/uuid and the string the ref points at — one value, by string equality. Every other clause omits its UUID. Expression refs work the same way but key off the expression's lib/expression-name string, so expressions rarely need an explicit lib/uuid.)

When you do need one, always mint it with mb uuid — never write, guess, or copy a UUID yourself. A hand-authored value is rejected pre-flight as not-a-v4 ("a1", "uuid-1", "agg-uuid-001" → must be a UUID v4 (RFC 4122) — run `mb uuid` ), or if it looks valid risks colliding with another clause. Mint just the few you reference (mb uuid --count 2 --json; this also covers native template-tag ids and any other format: "uuid" slot).

Authoring loop: print-schema → dry-run → run

mb query is the canonical authoring surface. Three modes:

mb query --print-schema --profile <n> > ./.scratch/mbql-schema.json   # 1. fetch the schema
mb query --file q.json --dry-run --profile <n>                  # 2. validate, no network
mb query --file q.json --profile <n> --json                     # 3. validate + run

• --print-schema emits { schema, defs } where defs carries id.yaml / parameter.yaml / ref.yaml / temporal_bucketing.yaml keyed by the path used in the schema's $refs. Read it first for any non-trivial query — cheaper than guess-and-fail.
• --dry-run validates and emits { ok, errors: [{ path, message }] }. Exit 0 valid, 2 invalid. No request sent. Iterate until ok: true.
• run (no flag) validates, then on success sends to /api/dataset. On validation failure it writes the same envelope, exits 2, and never sends.

path is a JSON Pointer into the body (/stages/0/aggregation/0); message is the validator error. Exit codes: 0 valid + ran, 2 validation failed / malformed body, 1 server-side error after a valid pre-flight.

Pre-flight is a lightweight shape check, not the full backend validator. It checks JSON shape, lib/uuid format, and enum values — not operator names, the first-stage source rule, or whether a reference resolves. A clean --dry-run is necessary but not sufficient: a body can pass pre-flight and still fail on the server (exit 1). The server is the authority — when a run fails, read its error and fix the body. Common ones:

• not a known MBQL clause → a misspelled or unsupported operator. Check the vocabulary in operators.md (mb skills get mbql --full).
• Initial MBQL stage must have either :source-table or :source-card → the first stage is missing its source (a numeric table or card id); only the first stage takes one, later stages read the previous stage's columns.
• Invalid :expression reference: no expression named "X" (or an invalid :aggregation reference) → a ref points to an expression name / aggregation lib/uuid that isn't defined in the query; fix the target string.
• Duplicate :lib/uuid → you reused a lib/uuid. Omit them (the server mints unique ones) or give each clause a distinct value.

A successful run emits the compact envelope by default: data.rows + slim data.cols (name, display_name, base_type, semantic_type). Pass --full for the raw /api/dataset envelope (results_metadata, native_form, per-column fingerprints/field_ref) only when you need that metadata; --fields data.rows narrows to rows alone. mb query also runs a native query — author it as an mbql.stage/native stage (pre-flight-validated like any MBQL body; see native-sql).

--skip-validate bypasses pre-flight and sends as-is — use only when the bundled schema disagrees with what the server actually accepts (drift / false negative). Mutually exclusive with --dry-run. Same flag exists on card create/update and transform create/update.

Where the query is consumed

The same body and pre-flight apply everywhere a query is embedded. Each pre-flights only when the value is the mbql/query shape (lib/type: "mbql/query"); legacy shapes skip it; --skip-validate bypasses.

Command	The query lives at	Notes
mb query	the whole body	ad-hoc run against /api/dataset
card create / card update	dataset_query	a flat mbql/query — see footgun below
transform create / transform update	source.query (when source.type is query)	materializes to a warehouse table
measure create / measure update	definition	exactly one aggregation, no filters
segment create / segment update	definition	filter macro tied to a table

Footgun: dataset_query is the flat mbql/query, not a legacy envelope

The most common mistake. The legacy shape { "type": "query", "database": N, "query": {…} } looks similar but is wrong. dataset_query (and source.query, and definition) is the mbql/query value itself:

"dataset_query": {
  "lib/type": "mbql/query",
  "database": 2,
  "stages": [{ "lib/type": "mbql.stage/mbql", "source-table": 190,
               "aggregation": [["count", {}]] }]
}

No type:"query" wrapper, no query: nesting. If you wrap the query inside a legacy envelope the CLI rejects it pre-send with a ConfigError (no --skip-validate gets it past). If it reached the server it would store silently and fail at run time with Initial MBQL stage must have either :source-table or :source-card.

Legacy formats you may encounter

Older Metabase servers used a different query envelope (sometimes called MBQL 4 / "legacy MBQL"); the mbql/query shape above is what recent servers store and return. You won't author the legacy shapes, but you may see them in queries created long ago. Anything not lib/type: "mbql/query" is sent as-is and normalized server-side — you lose validation, so don't author these:

• Legacy structured — { "type": "query", "database": N, "query": { "source-table": T, … } }
• Flat native — { "type": "native", "database": N, "native": { "query": "SELECT …" } } — the server accepts it, but author the native stage instead (native-sql).

mb query --file probe.json runs these directly; --dry-run on them returns { ok: true, errors: [] }. Don't author them by hand — build a legacy or complex query in the Metabase UI and pull the body with mb card get <id> --full --json / mb transform get <id> --full --json (which returns the mbql/query shape).

Joins and FK traversal

Two ways to read columns from a related table.

Explicit join. A stage's joins array holds join objects, each with three required keys: stages (the joined source as its own one-stage array carrying source-table/source-card), conditions (the ON clause — [op, {}, leftRef, rightRef], slot-1 rule and all), and alias (the string name you address joined columns by). Optional strategy (left-join default; also right-join / inner-join / full-join) and fields ("all" | "none" | an array of refs — which joined columns to select). Reference a joined column anywhere downstream by putting join-alias in the field options:

"joins": [
  {
    "alias": "Customers",
    "strategy": "left-join",
    "stages": [{ "lib/type": "mbql.stage/mbql", "source-table": 170 }],
    "conditions": [
      ["=", {}, ["field", {}, 1711], ["field", { "join-alias": "Customers" }, 1684]]
    ],
    "fields": "none"
  }
],
"breakout": [["field", { "join-alias": "Customers" }, 1682]]

Left ref is a column of the stage's own source (1711 = orders.customer_id); the right ref carries join-alias and points at the joined table's key (1684 = customers.id). Every later reference to a joined column (1682 = customers.plan) needs that same join-alias. Stack multiple objects in joins, each with its own alias.

Implicit FK join via source-field. For a single-hop FK lookup, skip the join — put the FK column's id in the target field's source-field option and Metabase traverses the relationship:

["field", { "source-field": 1711 }, 1682] // orders.customer_id → customers.plan

source-field is the FK field id (orders.customer_id); the third element is the target field id in the related table (customers.plan). Use it for "show a column from the table this FK points at"; reach for an explicit join when you need a non-FK condition, a non-default strategy, or control over which joined columns return.

Multi-stage pipelines

Stages run in order; each reads the previous stage's output columns — the breakouts and aggregations it produced — referenced by string name + base-type, not a numeric field id. Only the first stage takes a source-table/source-card. Add a stage to operate on an aggregate (you can't filter or order by an aggregation within the stage that computes it): aggregate, then filter the aggregate, then order + limit.

"stages": [
  { "lib/type": "mbql.stage/mbql", "source-table": 175,
    "aggregation": [["sum", { "name": "total" }, ["field", {}, 1715]]],
    "breakout": [["field", {}, 1711]] },
  { "lib/type": "mbql.stage/mbql",
    "filters": [[">", {}, ["field", { "base-type": "type/BigInteger" }, "total"], 0]] },
  { "lib/type": "mbql.stage/mbql",
    "order-by": [["desc", {}, ["field", { "base-type": "type/BigInteger" }, "total"]]],
    "limit": 3 }
]

Later stages address the first stage's aggregation by the name you gave it ("total") — set that name, or the column lands as the default sum / count / … and you reference that string instead.

Window functions sit in aggregation next to ordinary aggregates. offset reads a value from another breakout row — month-over-month is offset of a sum by -1 against a monthly breakout:

"aggregation": [
  ["sum", { "name": "revenue" }, ["field", {}, 1715]],
  ["offset", { "name": "prev_month" }, ["sum", {}, ["field", {}, 1715]], -1]
],
"breakout": [["field", { "temporal-unit": "month" }, 1717]]

Binning is a breakout-level field option — bucket a numeric column into ranges with ["field", { "binning": { "strategy": "num-bins", "num-bins": 5 } }, 1715], the numeric counterpart to the temporal-unit bucket. Strategies (num-bins / bin-width / default) are in the operator reference.

Naming aggregation output columns

Default aggregations materialize as count, count_where, avg, avg_2, sum, … — fine for an ad-hoc run, ugly for a transform target table or card column. Set name (the warehouse column name) and display-name (the UI header) in the aggregation's options:

["count", { "name": "shipments_shipped", "display-name": "Shipments shipped" }]

Operator reference

The full operator vocabulary — filter operators (=, !=, <, between, contains, is-null, …), aggregation functions (count, sum, avg, distinct, count-where, share, …), expression operators (arithmetic, string, temporal), temporal-bucketing units, and binning strategies — lives in this skill's references/operators.md, in numeric-id form. Load it on demand rather than dumping the schema:

mb skills get mbql --full     # appends references/operators.md to this body
mb skills path mbql           # → the skill dir; then Read references/operators.md

mb query --print-schema is the exhaustive-but-heavy fallback (the full JSON Schema, ~1600 lines). The cheat-sheet covers the vocabulary; the --dry-run loop settles any disagreement.