save_url

Fetch a web page, extract its content as Markdown, and save it to the knowledge base with automatic chunking, embedding, and optional memory extraction. Use this to capture web content for future search and retrieval.

Parameters

Parameter	Type	Required	Default	Description
`url`	string (URL)	REQUIRED	-	URL of the web page to save
`title`	string	OPTIONAL	auto-detected	Override title (max 500 chars; auto-detected from page if omitted)
`tags`	string[]	OPTIONAL	-	Optional tags for organization
`extractMemories`	boolean	OPTIONAL	`false`	Extract entities and facts from content and store as memories

Example Request

Basic Save

{
  "url": "https://example.com/blog/interesting-article"
}

Save with Tags and Custom Title

{
  "url": "https://example.com/docs/api-reference",
  "title": "Example API Reference",
  "tags": ["api", "reference", "example"]
}

Save with Memory Extraction

{
  "url": "https://example.com/team/alice-bio",
  "tags": ["people", "team"],
  "extractMemories": true
}

Response

{
  "success": true,
  "documentId": "abc-123-def-456",
  "title": "Example API Reference",
  "url": "https://example.com/docs/api-reference",
  "chunksCreated": 12,
  "contentLength": 8450,
  "memoryExtraction": {
    "entitiesFound": 3,
    "relationsFound": 2,
    "observationsCreated": 7,
    "observationsDuplicate": 1
  }
}

The memoryExtraction field is included only when extractMemories is true and extraction succeeds.

Error Responses

Error	Cause	Fix
`Database not configured`	Missing Supabase credentials	Set `SUPABASE_URL` and `SUPABASE_SERVICE_KEY`
`Embedding provider not configured`	Missing embedding API key	Set `OPENAI_API_KEY`, configure Ollama, or set `GOOGLE_AI_API_KEY`
`Only http and https URLs are supported`	Non-HTTP URL scheme	Use an `http://` or `https://` URL
`URL targets a private or internal address`	SSRF protection triggered	Use a publicly accessible URL
`Failed to fetch URL: HTTP {status}`	Remote server error	Verify the URL is accessible
`Unsupported content type`	Non-HTML response	Only HTML pages are supported
`Response too large`	Page exceeds 5 MB limit	Use a smaller page or save content manually via `add_note`
`No content extracted from the page`	Empty or unparseable page	Check the page has readable content

Notes

This tool writes data to your knowledge base — it creates a new document. It is not destructive (it does not overwrite existing documents).
The page content is automatically converted to Markdown, chunked, and embedded for semantic search.
When extractMemories is enabled, entities and facts are extracted from the content and stored in the memory graph (requires ENABLE_MEMORY).
If the title is not provided, it is auto-detected from the page's <title> tag or <h1> heading.
The saved document's sourceType will be url.

add_note - Add text content directly as a note
search - Search saved content
get_document - Retrieve a saved document by ID
extract_memories - Extract memories from an existing document

On this page