query_data

Use query_data to run read-only GraphQL queries against your tenant’s data: the same schema get_schema documents, with variables, size limits, and export for results too large to return inline.

Input

Field	Type	Required	Notes
query	string	Yes	A read-only GraphQL query, ≤ 64 KB. Anything starting with mutation or subscription is rejected.
variables	object	No	Variables referenced by the query.
export	object	No	Large-result export: { "format": "csv" } or { "format": "jsonl" }.

query_data requires the read:data scope.

Read-Only Enforcement

query_data only runs read operations: a query whose top-level operation is mutation or subscription is rejected before it executes. Use suggest_change to propose writes instead; those still route through human review before anything changes.

Examples

List item types. Get every item type’s slug and name in one call.

{
  "name": "query_data",
  "arguments": {
    "query": "query ListTypes { itemTypes { slug name } }"
  }
}

Paginated items. items returns a paginated object, not a bare list: read the nested items field for the rows.

{
  "name": "query_data",
  "arguments": {
    "query": "query OpenIssues($itemType: String!, $limit: Int) { items(itemType: $itemType, limit: $limit) { items { id title status } total page totalPages } }",
    "variables": { "itemType": "issue", "limit": 10 }
  }
}

total is the full matching count regardless of page size: prefer itemStats when you only need a count.

Counting without paging. Use itemStats for counts and status breakdowns without paging through items.

{
  "name": "query_data",
  "arguments": {
    "query": "query Counts($itemType: String) { itemStats(itemType: $itemType) { totalCount overdueCount byStatus { group label count } } }",
    "variables": { "itemType": "issue" }
  }
}

Large export. Add export when a result set is too large to return inline.

{
  "name": "query_data",
  "arguments": {
    "query": "query AllOpen { allItems(status: \"OPEN\") { id title status } }",
    "export": { "format": "csv" }
  }
}

List workflow templates. Get every template’s name and description in one call. take is capped at 100: page with skip for larger libraries.

{
  "name": "query_data",
  "arguments": {
    "query": "query Templates { workflowTemplates(take: 100, isActive: true) { id name description itemTypeId isPublic isSystem } }"
  }
}

Full template detail. A template’s steps and flow map are stored as JSON on the record, not as a nested field selection: nodes is the array of steps (each node’s data carries its label, description, and any formData.fields form schema) and edges is the diagram map (source → target, with labels on branch edges).

{
  "name": "query_data",
  "arguments": {
    "query": "query Template($id: String!) { workflowTemplate(id: $id) { id name description nodes edges metadata } }",
    "variables": { "id": "TEMPLATE_ID" }
  }
}

nodes, edges, and metadata come back as raw JSON: parse them client-side to walk the steps and branching. (Templates hold steps as JSON; a running workflow instead exposes its steps as queryable rows via step, allSteps, and related fields.)

Large Results and Export

Tool responses are capped at 2 MB. If a result is close to or over that limit, narrow the field selection, page it with items/limit, or set export.

With export.format set, query_data writes the full result to a downloadable file. The response itself carries a sample of rows, a signed downloadUrl, and metadata such as the row count: not the full dataset inline. Download links expire; fetch a fresh one anytime with the dataExport(exportId) query. Exports also note whether the result was truncated.

Query Discovery

get_schema’s queryReference field returns the full catalog of read query fields and their signatures: every field you can put inside a query_data call. The same catalog is documented in full on GraphQL.

Agent Notes

• Narrow field selections to what you need.
• Prefer variables over string-interpolating values into the query.
• Use stats fields like itemStats for aggregates instead of paging through every item.
• Use suggest_change for writes: query_data never mutates data.