Integration

MCP Per Tool

Detailed parameter documentation for every MCP tool.

Complete reference for each MCP tool with parameters, examples, and response formats.

rag_search

Search your document index for relevant chunks. Supports semantic and hybrid (dense + BM25) search, document-metadata filtering, and an optional knowledge-graph subgraph (rendered as a PNG image).

Parameters

Parameter	Type	Default	Description
`query`required	`string`	—	Search query text
`k`	`integer`	—	Number of results to return (default: 10)
`threshold`	`float`	—	Min similarity score (0-1, default: 0.5)
`document_ids`	`list[string]`	—	Optional list of document IDs to filter search (default: null)
`metadata_filter`	`object`	—	Optional document-metadata key/value pairs to filter by; only documents matching every pair are searched (default: null)
`include_graph`	`boolean`	—	Also return the supporting knowledge-graph subgraph (default: false)
`render_graph_image`	`boolean`	—	When include_graph is true, render the subgraph as a PNG image (default: true)

Example Call

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "rag_search",
    "arguments": {
      "query": "What is the return policy?",
      "k": 5,
      "threshold": 0.7,
      "metadata_filter": { "type": "policy" },
      "include_graph": true
    }
  }
}

Example Response

Each result carries score_type (score semantics), optional component_scores (dense/BM25 inputs for hybrid results), and a clickable document_url. When include_graph is true and a subgraph is found, the response is a list of content blocks: an image (the rendered subgraph) plus a text block holding the JSON below.

{
  "query": "What is the return policy?",
  "results": [
    {
      "score": 0.92,
      "score_type": "rrf_fusion",
      "component_scores": { "dense": 0.82, "bm25": 12.4 },
      "document_id": "42",
      "document_name": "policies.pdf",
      "content": "Our return policy allows returns within 30 days...",
      "document_url": "https://app.guidedmind.ai/dashboard/rag-builder/7?documentId=42"
    }
  ],
  "count": 1,
  "search_time_ms": 234,
  "rerank_method": "rrf",
  "subgraph": { "nodes": [], "edges": [] },
  "project_url": "https://app.guidedmind.ai/dashboard/rag-builder/7"
}

For dense (semantic-only) search, score_type is cosine_similarity and component_scores is omitted. For hybrid search, score is a rank/fusion value, not a probability — use component_scores to inspect the dense and BM25 inputs.

upload_document

Upload a document to your project (processing happens separately).

Parameters

Parameter	Type	Default	Description
`file_path`required	`string`	—	Path to the document file
`file_name`required	`string`	—	Name of the file
`content_type`	`string`	—	MIME type of the file (default: application/octet-stream)
`config`	`object`	—	Optional processing config: preprocessing, chunking, embedding (default: null)
`metadata`	`object`	—	Optional metadata for the document (default: null)

Example Call

{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "tools/call",
  "params": {
    "name": "upload_document",
    "arguments": {
      "file_path": "./policies.pdf",
      "file_name": "policies.pdf",
      "content_type": "application/pdf"
    }
  }
}

Example Response

{
  "jsonrpc": "2.0",
  "id": 2,
  "result": {
    "document_id": "doc_xyz789",
    "filename": "policies.pdf",
    "status": "uploaded",
    "message": "Document uploaded successfully."
  }
}

upload_and_process

Upload a document and immediately start processing.

Parameters

Parameter	Type	Default	Description
`file_path`required	`string`	—	Path to the document file
`file_name`required	`string`	—	Name of the file
`content_type`	`string`	—	MIME type of the file (default: application/octet-stream)
`config`	`object`	—	Optional processing config: preprocessing, chunking, embedding (default: null)
`metadata`	`object`	—	Optional metadata for the document (default: null)

Example Call

{
  "jsonrpc": "2.0",
  "id": 3,
  "method": "tools/call",
  "params": {
    "name": "upload_and_process",
    "arguments": {
      "file_path": "./policies.pdf",
      "file_name": "policies.pdf",
      "content_type": "application/pdf",
      "config": {
        "chunking": { "chunk_size": 512, "chunk_overlap": 50 }
      }
    }
  }
}

Example Response

{
  "jsonrpc": "2.0",
  "id": 3,
  "result": {
    "document_id": "doc_xyz789",
    "filename": "policies.pdf",
    "status": "processing",
    "chunks_created": 42,
    "message": "Document uploaded and processing started."
  }
}

Use upload_and_process for single documents. Use upload_document when batching multiple files before triggering processing.

project_info

Return the identity and capabilities of the RAG project bound to this MCP connection. Call this first to learn what you are connected to.

Parameters

No parameters (the project is resolved from the connection's credentials).

Example Response

{
  "id": 7,
  "name": "Invoice Processing",
  "description": "All company invoices and policies",
  "domain": "Finance",
  "use_case": "Invoice verification",
  "document_count": 42,
  "has_knowledge_graph": true,
  "app_url": "https://app.guidedmind.ai/dashboard/rag-builder/7"
}

list_documents

List all documents in the project, newest first.

Parameters

No parameters (the project is resolved from the connection's credentials).

Example Response

[
  {
    "id": 42,
    "file_name": "policies.pdf",
    "status": "embedded",
    "chunks": 128,
    "created_at": "2026-01-15T10:30:00",
    "metadata": { "type": "policy" },
    "url": "https://app.guidedmind.ai/dashboard/rag-builder/7?documentId=42"
  }
]

update_document

Replace a document's content (and optionally its settings/metadata) and reprocess it in place — no duplicate document is created.

Parameters

Parameter	Type	Default	Description
`document_id`required	`integer`	—	The id of the document to update (see list_documents)
`content`required	`string`	—	The new full text content (markdown or plain text)
`config`	`object`	—	Optional settings: embedding (model, enable_bm25, hybrid_rerank_method, contextual_retrieval) and chunking (chunk_size, chunk_overlap, chunk_min_size, chunk_max_size, method)
`metadata`	`object`	—	Optional replacement document metadata (e.g. {"type": "skill"})

Example Call

{
  "jsonrpc": "2.0",
  "id": 4,
  "method": "tools/call",
  "params": {
    "name": "update_document",
    "arguments": {
      "document_id": 42,
      "content": "# Return Policy\n\nReturns are accepted within 30 days...",
      "metadata": { "type": "policy" }
    }
  }
}

Example Response

{
  "id": 42,
  "file_name": "policies.md",
  "file_size": 1024,
  "status": "settings-review-needed",
  "message": "Document updated and queued for reprocessing. It will be searchable again once processing completes.",
  "url": "https://app.guidedmind.ai/dashboard/rag-builder/7?documentId=42"
}

retrieve_skills

Retrieve the skills relevant to a task and consolidate them into one bundle. Skills are documents tagged with metadata {"type": "skill"}.

Parameters

Parameter	Type	Default	Description
`task_context`required	`string`	—	A description of the task the skills should help with
`k`	`integer`	—	Number of matching skill documents to include (default: 5)
`threshold`	`float`	—	Min similarity score (0-1, default: 0.5)

Example Call

{
  "jsonrpc": "2.0",
  "id": 5,
  "method": "tools/call",
  "params": {
    "name": "retrieve_skills",
    "arguments": {
      "task_context": "How do I reconcile a batch of vendor invoices?"
    }
  }
}

Example Response

{
  "found": true,
  "skill_count": 2,
  "skills": [
    { "document_id": "10", "file_name": "invoice-reconciliation.md", "score": 0.88, "url": "https://app.guidedmind.ai/dashboard/rag-builder/7?documentId=10" }
  ],
  "bundle": "# Consolidated Skill Bundle\n\nTask context: How do I reconcile a batch of vendor invoices?\n\n..."
}

reprocess_all

Queue every processed document in the project for reprocessing.

Parameters

Parameter	Type	Default	Description
`scope`	`string`	—	What to reprocess: full (default), embeddings, bm25, or graph

Example Call

{
  "jsonrpc": "2.0",
  "id": 6,
  "method": "tools/call",
  "params": {
    "name": "reprocess_all",
    "arguments": { "scope": "graph" }
  }
}

Example Response

{
  "queued": 42,
  "scope": "graph",
  "message": "Queued 42 document(s) for reprocessing (scope: graph)."
}

Use scope: "graph" after saving milestones/decisions to rebuild the knowledge graph so the new content flows into it.

← PreviousMCP Overview Next →n8n Integration