MCP tool reference

fabric_search

Find or ask Neural Fabric. When mode is omitted, search defaults to match, which retrieves stored entities and returns ranked matches. Answer generation requires explicit mode: "answer". Answer mode uses a model to compose a response, with confidence and supporting evidence.

{
  "query": "SO-03242607",
  "entityType": "SalesOrder",
  "detail": true,
  "limit": 5
}

{
  "query": "EXAMPLE_ORDER_ID",
  "results": [
    {
      "id": "EXAMPLE_ENTITY_ID",
      "type": "SalesOrder",
      "name": "EXAMPLE_ORDER_ID",
      "score": 12.5,
      "snippet": "Illustrative matching entity text."
    }
  ],
  "layer": "global",
  "candidatesNarrowed": 8,
  "totalEntities": 25,
  "timingMs": 14
}

{
  "answer": "Illustrative item A, Illustrative item B",
  "confidence": 1,
  "mode": "list-synthesis",
  "layer": "global",
  "evidenceCount": 0,
  "refusal": null,
  "topEvidence": []
}

Errors: omitting query fails schema validation. An empty results array means no matches were returned for this query and its filters.

fabric_list

Inventory of saved entities. Filter by type, paginate, or request a full entity with entityId and detail: true. Use graph when the question is about connections.

{
  "entityId": "YOUR_SALES_ORDER_ENTITY_ID",
  "detail": true
}

{
  "entities": [
    {
      "id": "EXAMPLE_ENTITY_ID",
      "type": "SalesOrder",
      "name": "EXAMPLE_ORDER_ID",
      "propertyCount": 6,
      "deleted": false
    }
  ],
  "total": 1,
  "showing": 1,
  "nextOffset": null,
  "layer": "global",
  "entityType": "SalesOrder"
}

{
  "entity": {
    "id": "EXAMPLE_ENTITY_ID",
    "type": "SalesOrder",
    "name": "EXAMPLE_ORDER_ID",
    "properties": {
      "orderNumber": "EXAMPLE_ORDER_ID",
      "status": "open"
    },
    "frequencyMetadata": null,
    "deleted": false
  }
}

fabric_read_source

Read extracted text from a source document attached to an entity. Use offset and maxChars to page through the extracted text. Text files decode directly; text from PDF, DOCX, and XLSX files is extracted on demand. The response contains text, not the original file bytes.

{
  "entityId": "YOUR_SALES_ORDER_ENTITY_ID",
  "sourceDocumentIndex": 0
}

Errors: missing entityId fails validation. An entity with no attached source cannot return file text.

fabric_add

Save an entity. The provider configured on Connect is used to classify structured properties and document uploads. Relationships are discovered from shared property values, not created as a separate write. Structured properties and an attached sourceDocument are different inputs; see Attach a source file.

{
  "entityType": "SalesOrder",
  "name": "SO-03242607",
  "properties": {
    "orderNumber": "SO-03242607",
    "customerName": "Clearview Commercial Lighting",
    "quoteReference": "QT-03242607",
    "customerPo": "CCL-2026-R117",
    "orderDate": "2026-04-02",
    "amount": 510105,
    "status": "open"
  },
  "actor": "docs-sample"
}

Errors: missing entityType or properties fails validation. PDF without processPdfOverMcp: true returns guidance to use the in-app upload path or retry with the flag. Prefer Qora’s plus control → Add files for PDFs. Extra unknown fields are rejected (additionalProperties: false). Keep 60 or fewer top-level keys and strings under 3,500 characters.

fabric_edit

Edit one existing entity. Records field-level provenance, tracks source divergence, updates search visibility, and queues connection refresh. Max 25 edits.

set requires path and value. remove requires path. rename requires from and path. setName requires value. setSemanticRole requires property and role; role must be identity, reference, or descriptive. removeSemanticRole requires property.

{
  "entityId": "YOUR_SALES_ORDER_ENTITY_ID",
  "edits": [
    {
      "op": "set",
      "path": "/properties/status",
      "value": "released"
    }
  ],
  "reason": "Release sample sales order",
  "actor": "docs-sample"
}

Errors: missing entityId or edits, more than 25 edits, or an unknown field fails validation. Confirm the change with fabric_list detail or fabric_provenance.

fabric_graph

Read-only view of how entities connect. Relationships are discovered when entities are saved, not created through this tool.

{
  "mode": "root",
  "rootEntityId": "YOUR_SALES_ORDER_ENTITY_ID",
  "includeExplanations": true,
  "depth": 2
}

fabric_provenance

History for an entity or relation. The trail is append-only. Provide entityId or relationId.

{
  "entityId": "YOUR_SALES_ORDER_ENTITY_ID"
}

The schema does not mark either ID required in JSON Schema, but the tool description requires one of them. Calls with neither are not useful.

fabric_delete

fabric_delete permanently removes an entity and its attached source documents after writing delete provenance.

{
  "entityId": "YOUR_SALES_ORDER_ENTITY_ID",
  "reason": "Remove sample data",
  "actor": "docs-sample"
}

Errors: missing entityId fails validation. After success, fabric_provenance with the same entityId still returns the recorded history, including the entity_delete entry. includeDeleted remains an inventory filter; it does not restore or return a hard-deleted entity.